COLLEGE STUDENTS’ ACHIEVEMENT AND UNDERSTANDING OF
EXPERIMENTAL AND THEORETICAL PROBABILITY: THE ROLE OF TASKS
By
Irini Papaieronymou

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Mathematics Education
2012

ABSTRACT
COLLEGE STUDENTS’ ACHIEVEMENT AND UNDERSTANDING OF
EXPERIMENTAL AND THEORETICAL PROBABILITY: THE ROLE OF TASKS
By
Irini Papaieronymou

This study examined the role of particular tasks implemented through two instructional
methods on college students’ achievement and understanding of probability. A mixed methods
design that utilized a pre-test and post-test was used. This included treatment and control groups,
each comprised of students in three sections of an introductory statistics course taught by the
researcher at a college in Cyprus. During the study, students in the treatment group worked in
small groups on four in-class activities about experimental and theoretical probability
(Instructional Method B), and students in the control group worked in small groups on solutions
to four sets of probability problems from the course textbook (Instructional Method A).
An initial analysis of pre-test scores indicated that the students in the control group had
comparable initial probability knowledge to the students in the treatment group. Quantitative as
well as qualitative analysis were then carried out to address the research questions. With regards
to students’ achievement on the multiple-choice items on probability, the results of the Wilcoxon
Signed-Ranks test which was carried out to analyze gain scores indicated that the multiplechoice scores of students in the control group were significantly lower on the post-test compared
to the pre-test. In the case of the treatment group, student scores on the multiple-choice items did
not increase significantly from the pre-test to the post-test. Possible explanations to this
phenomenon are provided in the last chapter of this dissertation. In addition, an analysis of

normalized gain scores was carried out. Positive as well as negative normalized gains existed in
both groups. The Mann-Whitney test resulted in a p-value of 0.001 (< 0.05) indicating that the
normalized gain scores of the treatment group were significantly different from the normalized
gain scores of the control group. Relative to students’ achievement on the open-ended items
included on the post-test, the Mann-Whitney test resulted in a p-value of 0.001 indicating that the
scores of the treatment group on these items were significantly higher than the scores of the
control group. Therefore, Instructional Method B was successful in producing significantly better
achievement scores than Instructional Method A.
In this dissertation, students’ understanding of probability was measured through a distractor
analysis and a qualitative analysis of audio-taped student conversations. Specific to the distractor
analysis relative to probabilistic heuristics defined in the literature, the mean percentage of
students who applied these mostly increased in the case of the control group whereas it mostly
decreased in the case of the treatment group. These results are in line with past research that
indicated that the use of activity-based instruction may help students with respect to probabilistic
misconceptions (e.g. Shaughnessy, 1977, 1981).
For the purposes of qualitative analysis, the framework by Jones, Thornton, Langrall and
Tarr (1999) was used. Based on the results of the qualitative analysis of students’ levels of
reasoning relative to the constructs presented on the framework i) Instructional Method B
(treatment) produced better results than Instructional Method A (control) relative to experimental
probability and the concept of sample space and ii) it was difficult to identify which instructional
method had a better effect on students’ understanding of theoretical probability, conditional
probability and independence, discrete probability distributions and the binomial distribution.

Copyright by
IRINI PAPAIERONYMOU
2012

To my family, for their support throughout the years
And most of all
To my beloved husband George, for all his love, patience and understanding

v

ACKNOWLEDGMENTS

In my pursuit to carry out this dissertation in my home country Cyprus, while being
registered as a PhD student at Michigan State University, USA I faced many challenges which
were overcome with the help of many people.
First, I would like to express my gratitude to my wonderful advisor, Dr Sharon Senk who
supported and guided me throughout my graduate studies in mathematics education. Dr Senk’s
thoughtful feedback, invaluable insights, challenging questions, and time devoted to emailing
and holding skype meetings with me, helped tremendously in completing each step of this
dissertation and keeping me on track to graduation. She has been instrumental in helping me
become a researcher and a more reflective mathematics instructor, and for these I will forever be
indebted!
Special thanks go to the rest of my committee members. Dr Jack Smith’s thoughtprovoking emails and personal conversations pointed me to useful sources of literature and
helped me in the formulation of a definition of the term ‘understanding’ in this study. The
statistical expertise of Dr Jennifer Kaplan and Dr Vincent Melfi, along with their feedback,
helped me with the methodological and quantitative analyses aspects of this study. Thank you all
for the time you devoted to this dissertation!
This study certainly would not have been possible without the students who volunteered
to participate. Thank you for your cooperation and willingness to be part of my research!
Moreover, special thanks go to Lisa Keller, Margaret Iding and Jean Beland for their prompt

vi

emails whenever I needed them and for their help in overcoming any administrative issues that
arose during the PhD process!
While working on this dissertation, I was lucky to meet a wonderful and kind-hearted
man, George, who I am happy and proud to be able to call my husband! His constant caring,
love, patience, sense of humor, and … wedding proposal, were sources of inspiration during the
writing process of the dissertation. Thank you for standing by me, for being my friend and
lifetime companion, and for traveling to the USA with me!
A deep thank you goes to my parents who supported me at every step of the way during
my studies and made my stay in the USA possible. My gratitude also goes to my uncle Petros,
for his guidance throughout my life, and my aunt Pitsa for her support and constant care. Thank
you to my wonderful brother, Michael whose love and support have always accompanied me in
life! I would also like to thank my grandparents, Christos and Irini, for instilling in me the
importance of education and a love for learning. Had they been alive I know that they would
have been proud of my accomplishments.
Special thanks go to my best friend Georgia who always stood by me. Thank you for the
countless hours you spent listening to me, for the myriad times you made me laugh, and for your
reassurance and belief in me whenever I reached an obstacle in my way! My gratitude also goes
to my dear friend Beste who supported me in my times of need and helped me not feel lonely
while at Michigan State University. You will always be in my heart! My sincerest thanks also go
to my friend Electra and her parents, uncle Nicos and auntie Anna, who had been my family
while in Michigan! Moreover, I would like to thank my friends Elena, Tina, Costas, Neelambari,
Carole, Nicole, Vicky, and Roselyn for making my life more enjoyable with their friendship!

vii

TABLE OF CONTENTS

LIST OF TABLES

xii

LIST OF FIGURES

xv

CHAPTER 1: INTRODUCTION
1.1 Importance of Probability
1.2 Teaching of Probability
1.3 Purpose of the Study
1.4 Research Questions
1.5 Definitions
1.5.1 Experimental Probability
1.5.2 Theoretical Probability
1.5.3 Probability Experiments with Real Data
1.5.4 Cooperative Learning
1.5.5 Achievement and Understanding
1.6 Overview of Methods
1.7 Overview of Chapters

10
10
11
12
14
16
17

CHAPTER 2: LITERATURE REVIEW
2.1 Probabilistic Reasoning and Understanding
2.1.1 Research on Heuristics
2.1.2 Research on Probabilistic Difficulties and Misconceptions
2.1.2.1 Law of Large Numbers
2.1.2.2 Other Basic Probability Concepts
2.1.2.3 Conditional Probability
2.1.2.4 Binomial Distribution
2.2 Probability Experiments with Real Data
2.3 Cooperative Learning
2.4 Understanding
2.4.1 Mathematical Understanding
2.4.2 Probabilistic Reasoning, Thinking and Understanding
2.5 Use of Understanding in this Study
2.6 Available Models
2.6.1 Shaughnessy (1992): Stochastic Understanding
2.6.2 Jones, Thornton, Langrall and Tarr (1999): Probabilistic Reasoning
2.7 Theoretical Framework for this Study
2.8 Summary

19
19
25
25
27
32
35
36
39
45
45
49
50
52
52
53
56
57

CHAPTER 3: CYPRUS EDUCATIONAL SYSTEM
3.1 Control of the Educational System
3.2 Language of Instruction in Schools in Cyprus

59
62

viii

1
5
9
9

3.3 Intended Curriculum
3.3.1 Probability in the Intended Secondary Curriculum
3.3.2 Probability at the Tertiary Level
3.4 Performance of Secondary School Students on Probability on TIMSS
3.5 Recent Developments in the Educational System of Cyprus
3.6 Research Site for this Study

63
63
65
66
67
71

CHAPTER 4: METHODS
4.1 Overview of the design
4.2 Research Site
4.2.1 College
4.2.2 Course
4.2.3 Timeline and Procedures
4.2.4 Pilot
4.3 Participants
4.4 Instruction
4.4.1 Treatment Group
4.4.2 Control Group
4.4.3 Commonalities between Treatment and Control Groups
4.5 Instruments
4.5.1 Student Background Questionnaire
4.5.2 Probability Pre-test
4.5.3 Probability Post-test
4.6 Data Analysis Plan
4.6.1 Quantitative Analysis
4.6.1.1 Initial Equivalence of Groups
4.6.2 Qualitative Analysis
4.7 Summary

73
75
76
79
81
82
83
88
89
92
92
92
94
95
95
100
105
110

CHAPTER 5: STUDENTS’ ACHIEVEMENT AND UNDERSTANDING OF
PROBABILITY – RESULTS OF QUANTITATIVE ANALYSIS
5.1 Effects of Instructional Treatment on Achievement
5.1.1 Comparison of (Normalized) Gain Scores
5.1.1.1 Comparison of Gain Scores Within Each Group
5.1.1.2 Comparison of Gain Scores Between Groups
5.1.2 Comparison of Scores on Additional Post-Test Items
5.1.3 Comparison of Post-Test Total Scores
5.2 Effects of Instructional Treatment on Understanding
5.2.1 Multiple-Choice Items: Content Assessed and Difficulty Level
5.2.2 Multiple-Choice Items: Distractor Analysis
5.3 Summary

111
111
111
119
121
124
125
126
130
133

ix

73

CHAPTER 6: STUDENTS’ UNDERSTANDING OF PROBABILITY – RESULTS OF
QUALITATIVE ANALYSIS
137
6.1 Information on Activities and Problem Sets
138
6.2 Inter-Rater Reliability
139
6.3 Effects of Instructional Treatment on Understanding of Basic Probability Concepts
6.3.1 Sample Space
141
6.3.2 Theoretical Probability of an Event
149
6.3.3 Experimental Probability of an Event
162
6.3.4 Summary
170
6.4 Effects of Instructional Treatment on Understanding of Conditional Probability and
Independence
6.4.1 Conditional Probability
171
6.4.2 Independence
177
6.4.3 Summary
179
6.5 Effects of Instructional Treatment on Understanding of Discrete Probability
Distributions
180
6.6 Effects of Instructional Treatment on Understanding of Binomial Distribution
184
6.7 How Results Address Research Questions
186
6.8 Summary
188
CHAPTER 7: SUMMARY, DISCUSSION AND RECOMMENDATIONS
7.1 Summary of the Study
7.1.1 Purpose
7.1.2 Methods
7.2 Summary of the Findings
7.2.1 Effects of Instructional Treatment on Students’ Achievement in
Probability
7.2.2 Effects of Instructional Treatment on Students’ Understanding of
Probability
7.3 Discussion of Findings
7.4 Strengths and Limitations
7.4.1 Strengths
7.4.2 Limitations
7.5 Implications and Recommendations for Further Research

193
194
194
197
197
199
206
212
214
215

APPENDIX A: MAJOR THEMES IN THE DOMAIN OF PROBABILITY

223

APPENDIX B: INSTRUMENTS

229

APPENDIX C: INSTRUCTIONAL MATERIALS

256

APPENDIX D: SCORING RUBRICS

298

APPENDIX E: STATISTICAL RESULTS

310
x

BIBLIOGRAPHY

314

xi

LIST OF TABLES

Table 2.1 Framework for Students’ Probabilistic Reasoning (Jones et al., 1999, p. 15)

54

Table 3.1 Probability Content in Grade 12 Mathematics Textbooks Used in Public Schools in
Cyprus
65
Table 3.2 Probability in the Reformed Intended Curriculum for Grade 10 in Cyprus

70

Table 3.3 Probability in the Reformed Intended Curriculum for Grade 11 in Cyprus

70

Table 3.4 Probability in the Reformed Intended Curriculum for Grade 12 in Cyprus

70

Table 4.1 Mathematics Course Requirements for Students at Research Site

74

Table 4.2 Number of Audio Recordings Collected in Each Group

91

Table 4.3 Pre-Test Items – Sources

94

Table 4.4 Content-Heuristic-Misconception Assessed By Each Multiple-Choice Item

98

Table 4.5 Descriptive Measures of Pre-Test Scores By Course Section

101

Table 4.6 Results of the Kruskal-Wallis Test for k-independent Samples

104

Table 4.7 Data Sources and Associated Measures

107

Table 4.8 Theoretical Probability: Levels of Reasoning and Examples of Representative Excerpts
108
Table 4.9 Data Sources and Associated Measures

110

Table 5.1 Changes in Descriptive Statistics from Pre-Test to Post-Test

112

Table 5.2 Wilcoxon Signed-Ranks Tests for Multiple-Choice Gain Scores

114

Table 5.3 Changes in Percent-Correct Responses on Multiple-Choice Items

119

Table 5.4 Descriptive Statistics for the Open-Ended Item Scores

122

Table 5.5 Descriptive Statistics for the Post-Test Total Scores

124

Table 5.6 Percent-Correct Responses on Multiple-Choice Items (N = 44)

127

xii

Table 5.7 Multiple-Choice Items with Positive Shift in Student Difficulty Level

128

Table 5.8 Multiple-Choice Items with Negative Shift in Student Difficulty Level

129

Table 5.9 Mean Percentage of Students Applying Particular Heuristics or Misconceptions

131

Table 5.10 Mean Percentage of Students Who Applied Particular Misconceptions

132

Table 6.1 Information on Activities and Problem Sets

139

Table 6.2 Percent-Agreement Between Coders of Transcript Excerpts

140

Table 6.3 Levels of Reasoning Exhibited by Students Relative to Sample Space

149

Table 6.4 Problem 4.9 (Levine, Krehbiel, and Berenson, 2010)

153

Table 6.5 Levels of Reasoning Exhibited by Students Relative to Theoretical Probability

162

Table 6.6 Levels of Reasoning Exhibited by Students Relative to Experimental Probability 170
Table 6.7 Problem 4.23 (Levine, Krehbiel, and Berenson, 2010)

175

Table 6.8 Levels of Reasoning Exhibited by Students Relative to Conditional Probability

177

Table 6.9 Levels of Reasoning Exhibited by Students Relative to Independence

179

Table 6.10 Activity 3: Distribution of Chips by Students and Explanation

182

Table B.1 English Language Proficiency

230

Table B.2 Pre-Test Item 8 – Marbles in Container

236

Table B.3 Pre-Test Item 9 – Marbles in Container

237

Table B.4 Pre-Test Item 10 Cross Tabulation

238

Table B.5 Post-Test Item 8 – Marbles in Container

247

Table B.6 Post-Test Item 9 – Marbles in Container

248

Table B.7 Post-Test Item 10 Cross Tabulation

249

Table B.8 Post-Test Item 16 Cross Tabulation

254

Table C.1 Activity 1 – Recording Results of Sum of Two Dice

259

xiii

Table C.2 Activity 2 – Results of Rolling Three Fair Dice

265

Table C.3 Activity 3 – Placement of Chips on Number Line

276

Table C.4 Activity 3 – Recording Result of Rolling Two Dice

278

Table C.5 Activity 3 – Probability Distribution for Sum of Two Dice

281

Table C.6 Activity 4 – Results of Free-Throw Basketball Attempts

287

Table C.7 Problem 4.8 Cross Tabulation

293

Table C.8 Problem 4.23 Cross Tabulation

295

Table C.9 Problem 5.3 Probability Distribution

296

Table D.1 Scoring Rubric for Post-Test Item 16a

298

Table D.2 Scoring Rubric for Post-Test Item 16b

299

Table D.3 Scoring Rubric for Items 16c) i), 16c) ii), and 16c) iii)

300

Table D.4 Scoring Rubric for Item 16c) iv)

301

Table D.5 Scoring Rubric for Items 16d and 16e

302

Table D.6 Scoring Rubric for Item 16f

303

Table D.7 Scoring Rubric for Item 17a

304

Table D.8 Scoring Rubric for Item 17b

305

Table D.9 Scoring Rubric for Item 17c

306

Table D.10 Scoring Rubric for Item 18a

307

Table D.11 Scoring Rubric for Items 18b, 18c, and 18d

308

Table D.12 Scoring Rubric for Item 18e

309

Table E.1 Percent-Correct Responses on Pre-Test and Post-Test Multiple-Choice Items

310

Table E.2 Results of Distractor Analysis of Multiple-Choice Items (Correct Response in bold)
311
xiv

LIST OF FIGURES

Figure 4.1 Pre-Test Score Frequencies By Course Section

102

Figure 5.1 Control Group % Correct Responses on Pre/Post-Test Multiple-Choice Items

116

Figure 5.2 Treatment Group % Correct Responses on Pre/Post-Test Multiple-Choice Items

118

Figure 7.1 Pre/Post-Test Item 11 Coin Figure

208

Figure 7.2 Pre/Post-Test Item 13 – Spinner

209

Figure B.1 Pre-Test Items 4 and 5 – Sample Space for Rolling Two Fair Dice

233

Figure B.2 Pre-Test Item 11 Coin Figure

239

Figure B.3 Pre-Test Item 13 – Spinner

240

Figure B.4 Pre-Test Item 14 – Ticket Boxes

241

Figure B.5 Post-Test Items 4 and 5 – Sample Space for Rolling Two Fair Dice

244

Figure B.6 Post-Test Item 11 Coin Figure

250

Figure B.7 Post-Test Item 13 – Spinner

251

Figure B.8 Post-Test Item 14 – Ticket Boxes

252

Figure C.1 Activity 1 – Sample Space for Rolling Two Fair Dice

260

Figure C.2 Activity 3 – Sample Space for Rolling Two Fair Dice

280

xv

CHAPTER 1
INTRODUCTION
'Tis all a checker-board of nights and days
Where destiny with men for pieces plays
Hither and thither moves, and mates, and slays,
And one by one back in the closet lays.
(Edward Fitzgerald’s translation of The Rubáiyát of Omar Khayyám, 1879; as cited in Woolfson,
2008, p. 191).
1.1 Importance of Probability
As the Rubáiyát notes, in life we are moved “hither and thither”; that is, life is
unpredictable, consisting of “a series of chance happenings that toss you this way and that,
sometimes for good and sometimes for ill” (Woolfson, 2008, p. 191). In 1998 people in Great
Britain were astounded by the news of the death of the son of a couple in Cheshire. The boy,
Harry, was aged eight weeks at the time of his death. The astonishment came about due to the
fact that a little more than a year prior to Harry’s death, the same couple lost their son
Christopher, aged eleven weeks at that time. In response to this sequence of events, the couple
was arrested and after a trial that hit headline news, the mother, Sally Clark, was charged with
the murder of her two sons.
What was the evidence that led to the mother’s arrest? Sally Clark insisted that the cause
of her sons’ deaths was SIDS (Sudden Infant Death Syndrome). Recent research has revealed
that the cause of SIDS is the existence of a brain abnormality in infant victims. This abnormality
prevents infants from sensing high carbon dioxide and low oxygen levels thus increasing their
risk of inhaling their own exhaled breath (Hawkes, 2006). However, these research results were
not known at the time of Sally Clark’s trial.
1

Pediatric consultant Sir Roy Meadow acted as an expert witness for the prosecution in the
Sally Clark trial and although he was principally supposed to provide medical evidence, he also
made a statistical statement that led the court to its decision to charge the mother with murder
(Batt, 2004). The statement by Meadow indicated that the probability of two children in the same
family dying of SIDS was 1 in 73 million. This probability sent Sally Clark to jail to serve two
life sentences.
The court’s decision caused an incorrect belief to arise: that the probability of a very rare
event occurring is the same as the probability that the defendant is innocent, also known as the
Prosecutor’s Fallacy (Kaplan & Kaplan, 2006, p. 191). Later on, this flaw in reasoning was
recognized, with the Royal Statistical Society protesting against Meadow’s claim through a
public announcement. Along with medical records - that were not available at the first trial - that
the first son had indeed died from a respiratory infection, as well as records that Sally Clark was
not an abusive or uncaring mother, the woman was eventually released from prison in 2003 and
Sir Roy Meadow was taken off the medical rolls.
The above unpleasant event indicates the role of probability in courts. Yet, probability is
encountered in many other aspects of everyday life. Probabilistic claims are made by weather
forecasters, physicians, journalists, and election polls. In order to be a well-informed citizen one
needs to understand the language and basic ideas of probability (Gal, 2004; Scheaffer, Watkins,
& Landwehr, 1998; Utts, 2003) and how probability can be used to model our world (Moore,
1997; Penas, 1987).
Given these assertions about the role of probability in everyday life, it is not surprising
that since the late 1950s there has been a strong call for an increase in the inclusion of probability
in mathematics curricula. In Europe, 1958 was marked by a study of mathematics education of
2

the member countries of the Organization for European Economic Cooperation (OEEC), which
in turn led to an international conference the following year in France. One of the conference
conclusions was that statistics and probability should be introduced in the curriculum
(Exarchakos, 1988).
Following this, the 1960s were also marked by suggestions for the inclusion of
probability in school curricula. In Europe, and in particular in Great Britain, a reform project was
carried out during this decade, namely the School Mathematics Project, which was aimed at
middle and high school students (Lordou-Kaspari, 2003). The project writers were concerned
with the gap between school-level and university-level mathematics as well as with the absence
of applications in school mathematics; so, they included statistics and probability in school
syllabi. In the US, a group of mathematicians and National Science Foundation (NSF)
representatives published Goals for School Mathematics in 1963, in which the importance of
“some ‘feeling’ for probability” for all students was indicated (Jones, 1970, p. 291).
Subsequently, in the US, the National Council of Supervisors of Mathematics (NCSM)
identified probability as one of the basic skills that students should acquire (1977). Furthermore,
in 1983 the National Commission for Excellence in Education (NCEE) published A Nation at
Risk, a report aimed at pointing out the immediate need for reform in education. In the report, the
NCEE suggested that high school graduates should understand and be able to apply elementary
probability and statistics. In the context of this new reform movement in the USA, the University
of Chicago School Mathematics Project (UCSMP) initiated the development of a comprehensive
set of curriculum materials for grades K-12 starting in 1983. Among the goals of UCSMP was
“to give increased emphasis to “newer” mathematics, especially statistics and probability”
(Lordou-Kaspari, 2003, p. 117).
3

It should be noted that in the early 1980s, the results of the Second International
Mathematics Study (SIMS) revealed that in grade 8, probability was not considered important or
was not taught at all, both in US schools as well as in schools internationally (Travers and
Westbury, 1989). However, this has changed since then. On the Third International Mathematics
th

and Science Study 1999 (TIMSS), 67% of US 8 grade mathematics teachers indicated that they
spent at least one period teaching simple probabilities (TIMSS International Study Center, 2000).
A possible factor contributing to this increased attention on probability in US classrooms is the
set of recommendations set forth by US national professional organizations at the secondary
(NCTM, 1989, 2000) and post-secondary levels (MAA, 1998). In particular, the National
Council for Teachers of Mathematics (NCTM, 2000) specified in the Data Analysis and
Probability Standard that emphasis should be placed on the understanding and application of
basic probability concepts. According to the NCTM probability concepts should become
increasingly sophisticated as students move through the grades. Moreover, at the post-secondary
level, the Mathematical Association of America (MAA, 1998) and the American Mathematical
Association of Two-Year Colleges (Cohen, 1995) have also supported an increase in the
importance placed on probability.
With regards to Cyprus – where this study took place - the results of TIMSS 2007
th

revealed that at the 8 grade level only 3% of class time is devoted to data and chance (Mullis,
Martin and Foy, 2008). Topics relative to this domain are considered to be for the more able
students and as such, only 3% of Cypriot 8

th

graders receive formal instruction in this domain;

that is, this 3% of Cypriot students receive the 3% of instructional time devoted to data and
chance.
4

Given the importance of probability in everyday life and its recommended increased
focus in school curricula, it is encouraging that the Ministry of Education and Culture in Cyprus
has recently initiated educational reform efforts which include revisions to the national
curriculum in grades K-12 (The Ministry of Education and Culture, Republic of Cyprus, 2008).
In 2010, reports were published by the Ministry of Education and Culture of Cyprus regarding
the intended curriculum for each school subject, including mathematics. This report specifies the
content, procedures, applications and experiences that students are expected to acquire while in
school. Probability is introduced at the elementary school level with increased emphasis as
students move through the grades. The details of the mathematics report relative to the domain of
probability are provided in Chapter 3 of this dissertation.
1.2 Teaching of Probability
Over the past couple of decades, there have been various reform initiatives concerning the
content and means of instruction in mathematics classrooms at the college level (MAA, 1998). In
the US, the MAA (1998) recommends that lectures be replaced by more active teaching methods
in which group work is used and emphasis is placed on “making sense of inherently quantitative
situations, problem formulation and heuristics at the expense of mechanics” (Retrieved May 10,
2009 from http://www.maa.org/past/ ql/ql_toc.html).
In addition to initiatives taken in the mathematics education community, undergraduate
education research in the STEM fields (science, technology, engineering, and mathematics) has
raised concerns regarding the teaching and learning in these areas. Undergraduate education in
many STEM fields relies heavily on lectures which do not promote “genuine understanding”
needed for further studies and work in these fields (National Research Council, 2003; Baldwin,
2009, p. 10). Instead, teaching environments should be designed that promote the involvement of
5

all students through the use of instructional procedures such as cooperative learning, “which
affects the head and hand while simultaneously affecting the heart, thereby potentially reversing
the negative trends noted in higher education” (Smith, Douglas, and Cox, 2009, p. 21).
Specific to the area of statistics and probability, many researchers have recommended
that there be a change in the way statistics courses are taught (Chance, 1997; Garfield 1994;
Shaughnessy, 1981; Watts, 1991; see also Keeler & Steinhorst, 2001). “One area that has
received less focus in this literature is the teaching of probability” (Keeler & Steinhorst, 2001,
Retrieved April 24, 2009 from http://www.amstat.org/publications/jse/v9n3/keeler.html). Much
has been written about people’s misconceptions and use of heuristics regarding judgment under
uncertainty but there is a lack of research on solutions to this phenomenon (Keeler & Steinhorst,
2001). People hold preconceptions about probability so instruction needs to be changed and be
carried out in such a way as to facilitate learning. In 1988, Garfield and Ahlgren recommended
that more research be carried out on teaching probability in ways that aid people overcome their
misconceptions whereas in his 1992 review of research on probability, Shaughnessy called for
research to be carried out on instruction related to students’ conceptual knowledge of probability.
Relative to these issues, the NCTM (2000) promotes the use of manipulatives and
emphasizes the study of real-world problems and their connections to data analysis and
probability. Moreover, several researchers suggest that instruction allows students to build
models and develop their thinking capability (Pollak, 1968; Klamkin, 1968, Fitzgerald, 1975; see
also Shaughnessy, 1977). To this end, it is recommended that statistics instruction relies less on
lecturing and more on active learning that uses group problem-solving, activities and discussions
(Cobb, 2000). Given that curriculum reform has brought data handling to the forefront, less
emphasis should be placed on “formal” probability and “an empirical frequency-based approach
6

to probability that is also an important foundation for later work in theoretical probability”
should be used (Watson, 2006, p.127).
Since Shaughnessy’s 1992 review of research on probability and statistics, a considerable
number of studies have been carried out on the teaching and learning of probability (Jones,
Langrall & Mooney, 2007). However, most of this research relates to students’ thinking of
probability with a minimal amount having been carried out relative to instructional methods
(Hirsch & O’Donnell, 2001). In addition, Shaughnessy (1992) indicated a lack of research in
probability learning and teaching outside of western countries. Since then, very few studies have
been found outside the western culture and only one was cited in a recent review by Jones,
Langrall & Mooney (2007) which was carried out in Cyprus that examined elementary school
students’ probabilistic thinking. Furthermore, Jones, Langrall & Mooney (2007) indicated that
“there has been little research on students’ thinking about experimental probability and even less
on students’ understanding of the connections between theoretical and experimental probability”
(p. 946).
During the three academic years spanning 2008-2011, the researcher of this dissertation
study taught mathematics and statistics at a private college in Cyprus. In the courses that students
take at this college, emphasis is placed on application problems in the area of business since the
college is a specialized business school. During the introductory statistics course offered at the
college and which the researcher first taught in spring 2009, these application problems were the
most challenging for students. Students at this college tend to be mostly native Cypriots whose
first language is Modern Greek. However, the medium of teaching is English which poses
difficulties for several students. Moreover, many of these students were weak in probability and
found it difficult to think critically and reason through probability problems. During
7

conversations with another mathematics instructor at the college, the researcher was informed
that students in the introductory statistics course in previous years carried similar weaknesses
and attitudes towards probability.
With regards to probability, it was not surprising to the researcher to come across student
weaknesses. The national curriculum in public secondary schools in Cyprus includes minimal
exposure to probability which occurs towards the end of grade 12 (Ignatiou and Zotos, 2003).
Given that this introductory statistics course (Statistics I) provides the first substantial amount of
exposure that these students have to probability, and that along with Statistics II, may be the only
formal statistics courses that the students at this college take that include material on probability,
it is important that these courses be taught in the most effective way.
Since these students are a vital element of the future workforce of Cyprus, it is important
to understand their thinking capabilities and improve their critical thinking skills. For these to
occur, instruction needs to provide students with opportunities to express their thinking and
reasoning through ways that motivate them to do so. Studies have shown that college students
“do not think critically and reflectively about important societal issues” but “courses in statistical
thinking have the potential to improve students’ general reasoning capabilities” (Derry et al.,
2000, p. 748). Students “need opportunities on a regular basis to engage with tasks that lead to
deeper, more generative understandings about the nature of mathematical concepts, processes,
and relationships” (Stein, Smith, Henningsen, & Silver, 2000, p.15; as cited in Jones, 2004, p.3).
Therefore, the statistics courses that the students at this college are required to take could be
taught in such a way as to foster their critical thinking capabilities. Perhaps the use of in-class
group activities that promote students’ reasoning about probability might have a positive effect
on students’ achievement and understanding of probability.
8

1.3 Purpose of the Study
With the above issues under consideration, the purpose of this study was to examine the
role of particular tasks on college students’ achievement and understanding of probability. In
particular, the study aimed to examine the potential benefits of performing probability
experiments that generate real data and completing probability activities on college students’
achievement and understanding of experimental and theoretical probability. The study took place
in a college-level introductory statistics course and examined the effects of:
i.

an instructional method that combined lectures and small-group cooperative learning
sessions during which students solved probability problems (control group) and

ii.

an instructional method that combined lectures and small-group cooperative learning
sessions during which students completed activities involving probability experiments that
generate real data (treatment group).

1.4 Research Questions
The research questions addressed in this study involve the use of two instructional
methods:
Instructional Method A: Using lectures and small-group cooperative learning sessions during
which students solve probability problems and
Instructional Method B: Using lectures and small-group cooperative learning sessions during
which students use activities involving probability experiments that generate real data to make
connections between experimental and theoretical probability.
Given Instructional Method A and Instructional Method B

9

1) What are the effects of using each of these instructional methods on college students’
achievement on probability and on their understanding of experimental and theoretical
probability?
2) Does Instructional Method B have a better effect on college students’ achievement on
probability and on their understanding of experimental and theoretical probability than
Instructional Method A?
1.5 Definitions
1.5.1 Experimental Probability
Experimental or empirical probability involves first the collection of data through
experiments or simulations (Jones, Thornton, Langrall & Tarr, 1999) and then the use of relative
frequency to provide a posteriori approximation to the probability of an event (Mojica, 2006).
According to this approach, probability is “the hypothetical number towards which the relative
frequency tends when stabilizing (Batanero, Henry, & Parzysz, 2005, p. 23; see also von Mises,
1928/1952). The modern generalization of this stabilization of the relative frequency after a large
number of trials is called the Law of Large Numbers.
1.5.2 Theoretical Probability
Theoretical probability involves the determination of the probability of an event through a
priori approach (Levine, Krehbiel, & Berenson, 2010); that is, before conducting any
experimental trials. A theoretical probability may be computed mathematically using numerical
or geometrical methods (Burdzy, 2009; Jones, 2004) and it is based on the idea of equally likely
outcomes (Aczel & Sounderpandian, 2002; Burdzy, 2009).
Laplace was the first who attempted to define probability using a mathematical
th

interpretation at the beginning of the 19 century. His interpretation was based on equally likely
10

events and so, excluded the possibility of unequally likely events. Laplace (1814/1995; see also
Burdzy, 2009, p. 16) provided a definition of probability as follows:
The theory of chance consists in reducing all the events of the same kind to a certain
number of cases equally possible, that is to say, to such as we may be equally undecided
about in regard to their existence, and in determining the number of cases favorable to the
event whose probability is sought. The ratio of this number to that of all the cases possible
is the measure of this probability, which is thus simply a fraction whose numerator is the
number of favorable cases and whose denominator is the number of all the cases possible.
1.5.3 Probability Experiments with Real Data
The statistics education community recommends that students in statistics classrooms
have access to and experiences collecting, analyzing and using real data (Hall and Rowell, 2008).
The American Statistical Association (ASA) (Aliaga et al., 2005; Franklin & Garfield, 2006)
indicates that real data comes in various forms: archival data, data generated in the classroom,
and data generated through simulations. According to Franklin and Garfield (2006) teachers may
use textbooks, journals, data repositories found on the web, data from a practicing research
scientist, as well as data generated through surveys or activities completed in class as sources of
real data.
Real data in this study was in the form of data generated through experiments (e.g. rolling
dice) carried out by students (treatment group) or in the form of archival data (treatment and
control groups). The course textbook (Levine, Krehbiel, and Berenson, 2010), which was used in
all sections of the introductory statistics course in which the study took place, provided real data
sets. Since the students in all sections of the course were given opportunities to work with
textbook problems and were assigned homework from the textbook, all participants were
provided with opportunities to work with real data. However, the students in the control group
only dealt with real data provided through archived data sets in the course textbook whereas the
11

students in the treatment group, in addition and prior to working with archived data sets,
performed experiments in which they generated real data themselves in small groups.
1.5.4 Cooperative Learning
Cooperative learning is defined as a “structured, systematic instructional strategy”
(Cooper & Mueck, 1990, p. 68) in which students are assigned specific roles and work towards a
common goal while being responsible for their own learning. Cooperative learning makes use of
small groups of two to ten students (Springer, Stanne, & Donovan, 1999) who work on
assignments, in-class activities or problems until all group members understand the material
(Gunawardena, 1998).
Research identified five elements of cooperative learning groups. First, such situations
are characterized by positive interdependence; students perceive that they must make a joint
effort and that each member has a unique contribution they can make to the group (Johnson &
Johnson, n.d.; Johnson & Johnson, 1985; Effandi and Zanaton, 2007). Positive interdependence
may be structured by asking group members to fulfill specific assigned roles, to agree on an
answer for the group or by giving the group a shared grade (Smith, Douglas and Cox, 2009).
A second element is that of individual and group accountability; the group is accountable
for achieving the set goal and each member is accountable for making a contribution to the group
(Johnson & Johnson, 1985). Individual accountability may be structured by giving individual
exams (Effandi and Zanaton, 2007). In order to achieve group accountability, the instructor may
provide feedback to the group regarding their performance on the activity or problem.
Third, cooperative learning promotes face-to-face interactions among students. The
instruction must ensure that students interact with one another in order to help each other

12

complete the task, explain orally to one another how to solve problems and discuss the strategies
used (Smith, Douglas and Cox, 2009).
The fourth component of cooperative learning groups is that of teamwork skills. Such
skills may be introduced by assigning students different roles in their groups (Smith, Douglas
and Cox, 2009). The last element of cooperative learning groups is that of group processing. In
structuring group processing, instructors should provide students with a specific (rather than
vague) task which is complex enough to warrant a group, and also provide sufficient time for
them to work in groups.
In this study cooperative learning was used as part of instruction in all course sections
(control and treatment). The small student groups were set up by the instructor based on criteria
that are described in Chapter 4: Methods of this dissertation. In the treatment group cooperative
learning was promoted through the use of small-group in-class activities. In the control group
cooperative learning was promoted through the use of small-group sessions during which
students solved probability problems. In all sections of the course, students were assigned
specific roles and received a shared grade for work completed in groups during class (positive
interdependence). Assigned roles aimed at promoting face-to-face interactions among group
members as they completed an activity or solved a problem. Groups received feedback on their
performance and a shared score for the group work submitted to the instructor (group
accountability) but were examined individually on the course final exam (individual
accountability). In addition, students were assigned different roles and these were rotated
between activities/problem sessions (teamwork skills). Moreover, activities were well-structured
(treatment group), problem sets were specifically selected from the course textbook (control
group) and students were provided with sufficient time to work on these (group processing).
13

1.5.5 Achievement and Understanding
Although an extensive literature can be found relative to mathematical understanding, a
definition of probabilistic understanding has not been explicitly established. Researchers defined
terms such as probabilistic reasoning and probabilistic thinking both of which entail the idea of
understanding. A detailed review of the mathematical meaning of understanding along with
definitions of probabilistic reasoning and thinking, as well as the available models for examining
students’ probabilistic reasoning are provided in Chapter 2 of this dissertation.
As Dewey (1933) indicated, “to understand is to grasp meaning” (p. 132) and this
meaning may be acquired by seeing something “in its relations to other things” (p. 137). In a
similar manner, Thompson and Saldanha (2003) defined understanding as “assigning meanings
according to a web of connections the person builds over time” (p. 99). However, Thompson
and Saldanha indicate that these connections are constructed through interactions that the person
has with “with his or her own interpretations of settings and through interactions with other
people” (p. 99). According to Skemp (1971), “to understand something means to assimilate it
into an appropriate schema” where a schema is “a conceptual structure” (as cited in Resnik &
Ford, 1981, p. 167). In this study, the extended meaning of understanding given by Thompson
and Saldanha (2003) of “assimilation to a scheme” is used; this allows for correct as well as
incorrect or inappropriate understandings (i.e. misconceptions) people may have.
In statistics education the term probabilistic reasoning is more prevalent as compared to
the term understanding. In particular, Jolliffe (2005) defined probabilistic reasoning as
“understanding and being able to explain and justify probabilistic processes” (p. 326). Garfield
(2003) defined reasoning about uncertainty as “[U]nderstanding and using ideas of randomness,
chance, likelihood to make judgments about uncertain events; knowing that not all outcomes are
14

equally likely; knowing how to determine the likelihood of different events using an appropriate
method” (p. 25).
The definition of understanding used in this study combines i) the definitions of
understanding provided by Dewey (1933) and Thompson and Saldanha (2003) and ii) the
definitions of probabilistic reasoning expressed by Garfield (2003) and Jolliffe (2005).
Therefore, students’ understanding of probability in this study means to grasp the meaning of
probability concepts by viewing them in relation to other concepts and to be able to explain the
processes used in the solution of probability problems or in the completion of probability
activities as students interact with one another in small groups. Such an understanding includes,
as Garfield stated, knowledge that not all outcomes have an equal probability of occurring and
the ability to use an appropriate method to determine the probability of an event. Students’
understanding of probability includes both appropriate as well as inappropriate understandings.
Therefore, two things needed to be specified for each item on the pre-test and post-test used in
this study with regards to probabilistic understanding:
i)

What is the goal of the item? That is, what probability content or concept is the item
addressing and the student trying to understand?
and

ii)

What has the student actually understood or what misconception does he/she hold relative
to the particular item?
In this dissertation, achievement is measured quantitatively by considering students’

responses to a probability pre-test and post-test. Understanding is measured in two ways: i) a
distractor analysis of student responses to the multiple-choice items on the pre-test and post-test
and ii) a qualitative analysis of audio-taped conversations as students worked in groups on
15

activities or problem sets. A distractor analysis, by itself, provides useful, but not sufficient,
information on students’ understanding since it fails to offer a complete view of what
understanding is. This is so because multiple-choice items do not provide access to a responder’s
reasoning and subsequently, to a responder’s explanation as to why he/she chose a particular
response. Without access to a responder’s reasoning, an incorrect response to a multiple-choice
item does not directly indicate the use of a particular misconception; it simply indicates the
“application” of a given misconception (J. P. Smith, personal communication, December 14,
2011). Therefore, both a distractor analysis and a qualitative analysis of student conversations
were carried out in the attempt to gain insight on students’ understanding of probability in this
study.
1.6 Overview of Methods
This dissertation took place in an introductory statistics course taught by the researcher in
spring 2010 at a college in Cyprus. The study aimed to examine the effects of two instructional
methods (as described in section 1.3) on college students’ achievement and understanding of
experimental and theoretical probability. These two methods were used during instruction on the
probability component of the course.
Participants included 44 college students split among three sections of the course. The
majority of these students used Modern Greek as their first language and English as their second
language. The medium of instruction in the course was English. Two of the sections served as
the treatment group whereas one section served as the control group.
In order to address the research questions posed in this study, a mixed-methods design
was used. A pre-test and a post-test were given to participants. Students in the treatment group
worked in small groups on four in-class activities about experimental and theoretical probability,
16

and students in the control group worked in small groups on solutions to four sets of probability
problems. In each group student conversations were audio-recorded. These data sources allowed
for both quantitative and qualitative analyses of data. All of the activities used in the treatment
group had an experimental and a theoretical component and all included parts which required
students to compare experimental and theoretical probabilities (see Appendix C). This provided
the means for emphasizing the connection between theoretical and experimental probability for
this relation is not directly apparent to students (Jones, Thornton, Langrall and Tarr, 1999). Yet,
“[I]t is that reciprocal dynamic of theoretically computed probabilities and observed relative
frequencies that may best contribute to the development of efficient probabilistic intuition”
(Fischbein and Gazit, 1984, p. 3).
Specifically in this dissertation, achievement is measured quantitatively through the
performance of students on a probability pre-test and post-test, using statistical analysis methods
as described in Chapter 4. Understanding is measured through a distractor analysis of multiplechoice student responses to the pre-test and post-test items as well as through a qualitative
analysis of audio-taped student conversations as students worked in small groups on activities or
problem sets. The framework for the qualitative analysis of the audio tapes is provided in
Chapter 2 and the way in which it was used is presented in Chapter 6.
1.7 Overview of Chapters
The study is organized into seven chapters. Chapter 1: Introduction includes the rationale
for conducting this research study along with the research questions and definitions of terms
used. In Chapter 2: Literature Review a thorough review of the available literature on issues
related to this dissertation is given. Specifically, the following are reviewed: i) probability
misconceptions held by secondary school students, college students and adults; ii) the use of the
17

experimental approach and real data in teaching and learning probability; iii) the effects of
cooperative learning and iv) the literature on mathematical and probabilistic understanding. Also,
the framework used in this study is presented. Chapter 3: Cyprus Educational System provides
an overview of the Cyprus educational system since the study was conducted at a college in
Cyprus. Moreover, a detailed account is provided on educational reform efforts currently taking
place in Cyprus specific to the area of mathematics and to the teaching of probability. Chapter 4:
Methods provides a description of the research design, participants, procedures and instruments
used in the control and treatment groups as well as the types of analysis carried out. Chapter 5:
Students’ Achievement and Understanding of Probability – Results of Quantitative Analysis is a
presentation of the results obtained through the quantitative analysis of the data whereas Chapter
6: Students’ Understanding of Probability – Results of Qualitative Analysis is a presentation of
the results obtained through the qualitative analysis of the data. Last, Chapter 7: Discussion and
Conclusions presents a discussion of the results along with conclusions drawn from the study.
This last chapter includes the strengths and limitations of this dissertation as well as
recommendations for further research.

18

CHAPTER 2
LITERATURE REVIEW
This study is informed by various bodies of literature. First, an examination of the
available research on students’ probabilistic reasoning is included in this chapter. This section
focuses on the use of heuristics as well as the difficulties and misconceptions held by secondary
school students and adults. Since the participants in this study were beginning college students, it
was expected that they would use such heuristics, encounter such difficulties and carry
probability misconceptions as those listed in the literature. Second, a review of research
concerning the use of experiments that generate real data in mathematics and statistics
classrooms at the upper secondary and undergraduate levels is provided. This second section is
related to one of the instructional methods implemented in this dissertation. Third, the effects of
using cooperative learning are discussed since this learning approach was employed in this study.
Next, the literature on the mathematical meaning of understanding is reviewed along with related
terms defined in the probability literature since this study examined students’ understanding of
probability. Last, the theoretical framework used in this study is described.
2.1 Probabilistic Reasoning and Understanding
Although understanding is a term widely used and extensively defined in mathematics
education, in statistics education the term probabilistic reasoning is more prevalent. Some
research on probabilistic reasoning focused on school children (Green, 1982a; Shaughnessy,
1981, 1992, 2003; Watson & Moritz, 2002) whereas another line of research focused on the
reasoning of college students and adults (Kahneman, Slovic, & Tversky, 1982; Konold, 1989,
1993, 1995). This review focuses on the probabilistic reasoning – difficulties, misconceptions,
and use of heuristics - of secondary school students and adults.
19

People often fail to use the rules and methods learned in statistics courses when making
decisions under conditions of uncertainty thus resulting in misconceptions. “A misconception is
a student’s erroneous concept that produces a systematic pattern of errors” (Khazanov, 2008, p.
180). Such misconceptions act as obstacles in students’ ability to master probability concepts.
The literature reviewed below suggests that probability misconceptions are resistant to change
(Konold, 1995). The issue is that, when faced with a correct concept, one does not immediately
perform a “complete replacement” of the erroneous one with the correct one (Vosniadou and
Verschaffel, 2004; as cited in Khazanov, 2008, p. 182). The likely scenario is that the
misconception is attached to the correct concept, leading to student inconsistencies in
probabilistic reasoning (Konold, 1995).
At other times, people use “intuitive strategies (often referred to as heuristics) for dealing
informally with situations that might otherwise be formally analyzed in terms of probability”
(Pratt, 2000, p. 603). These heuristics may lead to quick and reasonable conclusions but may also
result in judgment that comes at odds with probability theory. Commonly used heuristics relative
to probability are discussed in detail in the literature review that follows.
2.1.1 Research on Heuristics
Research on probabilistic reasoning has been carried out by educators as well as by
psychologists such as Kahneman, Slovic, and Tversky (1982) who pursued an area of inquiry
labeled as judgment under uncertainty. Participants in the series of studies that these
psychologists completed were 1500 college and college-preparatory students. Under the line of
inquiry pursued by Kahneman, Slovic, and Tversky (1982), humans make decisions using
heuristics such as representativeness and availability in their effort to estimate the probability of
an event or to compare the probabilities of outcomes. For Kahneman and Tversky (1972),
20

heuristics are “strategies that statistically naïve people use to make probability estimates” (see
also Jones and Thornton, 2005, p. 74). These strategies can be helpful but may also lead to
misconceptions.
According to the representativeness heuristic, an event is probable to the extent that it is
representative of the population from which it is drawn or to the extent that it is characteristic of
the process that generates it. For example, people tend to think that the sequence of coin tosses
HTHTTH is more likely than HHHTTT because the latter does not appear to be random, and that
HTHTTH is also more likely than the sequence HHHTHH because the latter is not characteristic
of the fairness of coins (Kahneman et al., 1982). In particular, the results of Kahneman, Slovic
and Tversky’s (1982) experiment revealed that 82% of students indicated that the birth order
BGBBBB was less likely than GBGBBG. Using a situation involving coin tossing, Konold et al.
(1993) found that 38% of 7

th

graders and 33% of 11

th

graders used representativeness.

Moreover, research carried out by Batanero, Serrano and Garfield (1996) with 137 fourteen-yearold students who had not previously studied probability and with 130 eighteen-year-old students
who had received formal instruction on probability indicated that representativeness was widely
used among students in both groups. Furthermore, Fast (1997; see Carter, 2005) used a similar
question to that of Kahneman et al. involving coin tosses and found that one third of the
participating pre-service teachers did not give the correct answer. This was in agreement with
results of a study by Ulep (1990) with pre-service secondary mathematics teachers.
It has been noted though that as students age, the use of the representativeness
“misconception” decreases (Fischbein and Schnarch, 1997, p. 101). Moreover, research by
Hirsch and O’Donnell (2001; see Shay, 2008) with 263 college students provided evidence that
the use of representativeness decreased based on the number of statistics courses taken. In
21

particular, 37.5% of participants who had taken two or more statistics courses used this heuristic
(Shay, 2008).
The representativeness heuristic manifests itself in the gambler’s fallacy or negative
recency in which one believes that a particular outcome is due to occur because it has not done
so for a while. Under the negative recency effect the person believes “intuitively that the
alternating outcomes seem to better represent a random sequence”, just like “the gambler
believes the events will balance at the end” (Bamberger, 2003, p. 28). On the other hand, under
the positive recency a person predicts that an outcome that has occurred repeatedly in the past
will keep occurring (Jones and Thornton, 2005). Fischbein and Schnarch (1997) found that
negative recency decreased with age whereas positive recency remained stable with age.
The second judgmental heuristic, namely availability, states that people think the
probability of an event is based on how easily occurrences of similar events can be brought to
mind. For example, when people using this heuristic are asked to estimate the probability of
heart-attack among middle-aged people, they tend to base their estimate on instances of heartattack among their middle-aged acquaintances (Kahneman et al., 1982). So, a person who has
more acquaintances who are middle-aged and had a heart-attack would give a higher probability
estimate than someone who does not have as many acquaintances with these characteristics. In a
study carried out by Fischbein and Schnarch (1997) with 20 students in each of grades 5, 7, 9,
and 11 as well as with 18 undergraduate pre-service mathematics teachers, it was found that the
frequency of the availability “misconception” grew stronger with age (p. 102). Moreover, Ulep
(1990) found that the availability heuristic was widely employed by pre-service secondary
mathematics teachers.

22

A third heuristic accounted for in the work of Kahneman and Tversky was that of
adjustment and anchoring. Under this heuristic, people start from an initial value (an anchor),
adjust it according to information given in the problem and make an inadequate probability
estimate (Tversky and Kahneman, 1974; see Jones and Thornton, 2005). There are two instances
to adjustment and anchoring: the conjunction fallacy and the disjunction fallacy. For example,
when asked which has the bigger chance, two consecutive sixes on a die or a single six on the
next roll of a die, someone using the conjunction fallacy would say that the first one has the
bigger chance (Jones and Thornton, 2005). That is, people using the conjunction fallacy think
that a compound probability can be higher than the probability of each single event whereas the
probability law states that P( A  B)  min( P( A), P( B)) . In contrast, someone using the
disjunction fallacy would say that getting a single six on the next roll of a die has a bigger chance
than getting at least one six in three rolls. Fischbein and Schnarch (1997) found that the
conjunction fallacy was very strong among students up to grade 9 and tended to be less strong
among high school and college students. In a study carried out by Diaz and de la Fuente (2007)
with 414 students in a college introductory statistics course, the conjunction fallacy was observed
in 71% of the responses (p. 136).
Although the use of heuristics may result in misconceptions, it can also lead to reasonable
estimates that are correct enough for practical purposes (Hirsch & O’Donnell, 2001;
Shaughnessy 1992). For example, availability is helpful in matters of decision-making whereas
representativeness is central to statistics where the aim is for a random sample to be
representative of the parent population so that the results of a study may be generalizable
(Borovcnik, 1986; see Shaughnessy, 1992).

23

Our task as mathematics educators is to point out circumstances in which judgmental
heuristics can adversely affect people’s decisions, and to distinguish these from situations
in which such heuristics are helpful. We are obliged to point out this difference to our
students; it is not that there is something wrong with the way our students think, just that
they – and we – can carry the usefulness of heuristics too far (Shaughnessy, 1992, p.
479).
The above statements as well as arguments brought forward in the last decade by
Stanovich (1999) and Gigerenzer (1991a, 1991b; see Stanovich 1999) point to alternative
interpretations of what Kahneman, Slovic, and Tversky (1982) demonstrated as “systematic
irrationalities” in human cognition. In particular, Gigerenzer (1991b) criticized the use of
heuristics to account for errors because i) they are errors “only from a particular narrow and
challengeable interpretation of probability”; ii) the explanations the heuristics attempt to provide
are a “little more than redescriptions of the phenomena they are intended to explain”; and iii) the
heuristics “are largely undefined concepts and can post hoc be used to explain almost
everything” (p. 102).
Stanovich and West (2000; as cited in Stanovich and West, 2003) argue that “research in
the heuristics and biases tradition has not demonstrated human irrationality at all and that a
Panglosian position … which assumes perfect human rationality is the proper default position to
take”. According to the Panglosian position, human irrationality is impossible although human
behavior may, in some instances, deviate from a normative model. Such deviations in human
behavior may be explained by: i) performance errors which may be attributed to the subject’s
lack of attention, memory lapses, or other minor psychological malfunctions; ii) incorrect norm
application; the experimenter has implemented the wrong normative model so, the problem lies
with the experimenter rather than with the subject; and iii) alternative task construal; the subject

24

interprets the task differently than the experimenter and thus responds to a different problem
(Stanovich, 1999).
As Nickerson (2004) points out the use of heuristics has gains as well as pitfalls. As long
as the heuristics work well in the situation they are applied, they may require simpler
computations and only a modest effort in addressing a problem, however, precision and
consistency of results are lost.
2.1.2 Research on Probabilistic Difficulties and Misconceptions
2.1.2.1 Law of Large Numbers
Piaget and Inhelder (1975) support that by age 12 children are fully aware of the effect of
sample size and the role of the Law of Large Numbers. In the experiments used by these
researchers, children were shown a box made up of two equal-sized compartments, with a funnel
at the top middle part. Balls were dropped in the box and children were asked whether a large
sample or a small sample of balls would result in a uniform distribution. Children indicated that
the larger sample was more likely to generate such a distribution (see Sedlmeier and Gigerenzer,
1997). Piaget and Inhelder concluded that children possessed an intuitive understanding of the
Law of Large Numbers. In agreement with this, Evans and Pollard (1985) concluded that
Overall subjects did quite well as intuitive statisticians in that their judgments tended,
over the experiments as a whole, to move in the direction required by statistical theory as
the levels of mean difference, sample size and variability varied (p. 68-69).
An analysis of the data collected by Zimmerman (2002) from 23 students in a high school
Advanced Placement Statistics class indicated that participants were able to recognize the effect
of many trials on empirical probability. “They recognized that with enough trials, the probability
would fluctuate less and settle towards a theoretical value” (p. 167). In addition, use of a
graphing calculator had a considerable effect on students’ ability to reason about probability
25

simulations during instruction. The graphing calculator enabled students to perform many trials
quickly “which in turn led to a deeper understanding of the long-run effects to empirical
probability” (p. 168). These results were in agreement with those of Aspinwall and Tarr (2001)
who examined the effects of an instructional program that made use of manipulatives in carrying
out simulations. Similar to the Zimmerman (2002) study, subjects were able to recognize the
effect of repeated trials on the probability of an event and developed a conceptual understanding
of the Law of Large Numbers.
However, another group of studies supports a different view than the one presented
above. In a study carried out by Fischbein and Schnarch (1997) with students in grades 5, 7, 9,
11 and prospective teachers, participants were asked to respond to two problems relating to the
effect of sample size. The researchers indicated that the misconception that sample size is
irrelevant was prominent among responses and that this misconception grew with age.
Jones and Harris (1982) investigated university undergraduate students’ understanding of
the Law of Large Numbers by means of three different tasks. The first task was similar to that
used by Piaget and Inhelder (1975) and involved the distribution of marbles between two
compartments; the second was a set of written tasks similar to those used by Kahneman and
Tversky (1972); and the third asked students to compare samples of counters drawn from a
parent population (see Jones and Harris, 1982). Participants performed well on the Piagetian task
but tended to ignore the relevance of the Law of Large Numbers when faced with tasks similar to
those used by Kahneman and Tversky, thus confirming the results of the latter. However,
participants performed significantly better on the tasks similar to those of Kahneman and
Tversky if they had first been tested on the Piagetian task. “Therefore, failure on the Kahneman

26

and Tversky task is typically due, not to a complete lack of insight, but to a failure to apply that
insight to a relevant instance” (Jones and Harris, 1982, p. 487).
Little is known about the explanations behind these inconsistencies in research results
relative to people’s attendance to sample size. For instance, Bar-Hiller (1979) suggested that a
factor influencing the use of sample size is that people tend to attend to “relative sample size
(relative to the population size)” and not “absolute” sample size (see Sedlmeier and Gigerenzer,
1997, p. 13). However, the results of her study were inconsistent and therefore, did not entirely
support this claim.
Sedlmeier and Gigerenzer (1997) carried out a review of studies relative to the use of
sample size which led them to distinguish between two types of tasks which had rarely been
distinguished in the literature: frequency distribution tasks and sampling distribution tasks. “A
frequency distribution is a distribution of values from one sample” whereas a sampling
distribution is “a distribution of means from independent samples of fixed size, drawn from the
same population” (p. 4). In frequency distribution tasks people judge how well the sample mean
estimates the population mean; in sampling distribution tasks people deal with the variance of
sampling distributions. This distinction helped explain the inconsistencies in the results of these
studies. Sedlmeier and Gigerenzer (1997) noted that “the distinction between frequency and
sampling distribution tasks was shown to be a strong predictor of participants’ use of sample-size
information” (p. 8). Moreover, when asked to construct sampling distributions, people tend to
construct frequency distributions.
2.1.2.2 Other Basic Probability Concepts
Fischbein, Nello and Marino (1991) undertook a study in which 618 students aged 9-14
were asked to identify various types of events. The results indicated that students tended to
27

confuse ‘certain’ and ‘possible’ events because, as the researchers explained, “usually, one tends
to relate the notion of ‘certain’ to that of ‘uniqueness’” (p. 527). Similarly, Chan (1997; see Li,
2000) found that 35% of the 425 participants aged 16-18 confused ‘certain’ and ‘possible’
events. Based on questionnaire responses and student interviews he concluded that this confusion
might be due to students’ misconception that i) a ‘certain’ event is the same as a ‘possible’ event
since both types of events are indicative that something will happen or ii) each type of event
consists of a single outcome but the student was unable to think of an outcome that would
happen for sure in a given situation such as in rolling a fair die (Li, 2000).
The linguistic difficulties associated with the terminology used for probability events are
also highlighted in a study by Green (1982a) in which participants included 2,930 students aged
11-16. The results indicated that students tended to think of i) ‘very likely’ to mean ‘certain’ or
‘always happens’ and ii) ‘not very likely’ and ‘unlikely’ to mean the same as ‘impossible’ or
‘cannot happen’. Students who tended to think that ‘low probability’ means ‘impossibility’ were
surprised when an outcome with a low probability still occured in an experiment (Konold, 1988;
see Ulep, 1990).
A study carried out by Green (1979, 1982b) with 3,000 students aged 11-16, examined
students’ thinking on probability concepts such as randomness, sample space, most likely event,
compound events and independence. Students at all levels exhibited difficulties distinguishing
between random and nonrandom distributions as well as with items relating to sample space.
Also, students had problems with the multiplication principle. In addition, Green (1983) asked
participants to decide which one of Bag A – containing 3 black and 1 white counters – or Bag B
– containing 6 black and 2 white counters – gives a bigger chance of picking a black counter.
More than 50% of the students chose Bag B because the bag had more black counters. Green
28

concluded that students relied on the absolute size instead of the relative size of outcomes when
comparing the likelihood of events. Alternatively one could argue that the students did not
understand proportional reasoning.
Furthermore, in a study by LeCoutre (1992) with 600 students, participants were asked to
decide which of two colors of marbles is more likely to be picked up from a bag containing two
red and one black marble. Many indicated that both colors had the same chance of being picked
up. LeCoutre labeled the students’ belief that all outcomes are equally likely as the
“equiprobability” bias; a bias that in the particular study persisted among children of all ages
regardless of their exposure to probability. The results of a study carried out by Albert (2003)
also revealed that college students in an introductory statistics course made use of the
equiprobability bias. The study took place before the participants received any formal instruction
in probability so, the results were indicative of their prior knowledge. The equiprobability bias
has also been exhibited among students across various cultures (Batanero, Serrano, & Garfield,
1996; Jones, Langrall and Mooney, 2007). This bias could be explained by students’ commonly
held belief that in order for an experiment to be “fair”, all outcomes need to be equally likely
(Jacobs, 1999; Shaughnessy, 2003).
Yet, although students hold the “fairness” belief, they sometimes tend to believe that
although all numbers on a die have an equal chance of occurring, the number 6 is the least likely
to occur when rolling a die (Konold et al., 1993). When students in a study carried out by Green
(1983) were asked to choose a number on a fair die that would give them the bigger chance of
winning a prize, the majority chose middle numbers such as 3 or 4 and avoided numbers 1 and 6.
Green (1982b) found that students often consider experimental outcomes to be equally
probable even though relative frequencies obtained from actually carrying out the experiments
29

show differently. In support of this issue, Shaughnessy (1977) found that students carry out such
an assignment of probabilities even when they perform the experiment themselves. In his study,
students performed many trials of an experiment and concluded that outcomes cannot be equally
likely. However, when they were requested to provide a probability model for the experiment
they assigned an equal probability to each outcome, claiming that theoretically the outcomes are
equally likely and pointing towards the belief that “every probability model is a uniform model”
(Shaughnessy, 1977, p. 303). In general, Green (1982b) found that students assign equal
probabilities to outcomes when they should not, and at times they assign unequal probabilities to
outcomes that are indeed equally likely.
A substantial amount of research on probabilistic thinking among college students has
been carried out by Konold (1989, 1993, 1995). An interview study with undergraduate students
indicated that people reason according to the outcome approach in which the goal seems to be
predicting the outcome of the next single trial (Konold, 1989). For example, when students were
asked to interpret the meaning of “70% chance of rain today” they predicted that “[I]t’s going to
rain today” (p. 68). Individuals who use the outcome approach tend to interpret probability
values as follows: a 50% chance means a total lack of knowledge about the outcome thus leading
to a “don’t know” decision; a probability value sufficiently higher than 50% (close to 75%) leads
to a “yes” decision; and a probability value sufficiently lower than 50% leads to a “no” decision
(Konold 1989, 1995). If the probability is close to 0% students view the event as impossible and
if the probability is close to 100% they view the event as certain. Only when the probability is
close to 50% they view an event as random. Ulep (1990) also found that pre-service secondary
mathematics teachers made use of the outcome approach.

30

Related to the outcome approach and equiprobability bias is a misconception identified
by Rubel (2006), namely the 50/50 approach. One of the questions that the 173 male participants
in grades 5, 7, 9, and 11 were asked to respond to involved the probability of getting one head
and one tail when two coins are tossed. More than half of the students provided the correct
answer of ½. However, when asked to justify their response they used the outcome approach,
predicting the probability of getting heads or tails on a single toss of a coin. In a follow-up
interview with 33 of the participants, one of the questions asked for the probability of getting all
tails when three coins were tossed. One of the students who said 50% gave the justification that
“unless something affects the way the quarters come down, it’s still going to be equal” (p. 52).
Overall, 40% of participants used the 50/50 approach on at least two questions.
Students also tend to find the concepts of independence and mutual exclusivity difficult.
Based on the results of a survey of 219 participants enrolled in a college-level introductory
probability and statistics course, “students tend to impose an artificial separation between the
concepts of mutual exclusivity and independence” (Manage and Scariano, 2010, p. 16). The
survey was administered at the end of the semester, after students had received formal instruction
on the two concepts. A substantial percentage (68.3%) of them believed that if two events are
mutually exclusive then they are independent; that is, they held the misconception that
“independence” means “separation” (Manage and Scariano, 2010, p. 16). Also, 36.1% of the
students thought that if two events are independent then they are mutually exclusive. According
to the researchers, the reason for the decrease in the percentage (68.3% to 36.1%) is that when
students come across the term “mutually exclusive” they immediately think of “separation” as
“independence” whereas when faced with the term “independence” no such immediate intuition
comes to mind.
31

Antoine (2000) also found that the 66 liberal arts students in a finite mathematics course
tended to confuse independence with mutual exclusivity and vice versa, and had difficulty
providing examples illustrating each of these types of events. When students were asked to
indicate whether events were independent based on given probability values, they tended to
apply previously learned definitions incorrectly into formulas. For example, some students
considered the union and intersection of events to be the same. Moreover, students focused on
key words which had a non-mathematical meaning. In one of the items two events were defined
as: “Event A: Getting at least one tail and Event B: Exactly three tails” (p. 61). A typical
response to whether these were independent was: “Events are dependent because the word “tails”
can be found in the intersection (overlap) of events A and B” (p. 61). An issue that came up was
that arithmetic and algebra skills played a role in students’ ability to solve problems on
independence since many of them incorrectly used principles of multiplying with exponents and
checking the equality of equations.
2.1.2.3 Conditional Probability
Research was also carried out with regards to students’ reasoning about conditional
probability (Pollatsek, Well, Konold & Hardiman, 1987; Watson & Moritz, 2002). Watson and
Moritz (2002) investigated probability or frequency estimates that students in grades 5-11 give
for a conditional event and its inverse. For example, their research included items such as the
following:
Please estimate:
(a) Out of 100 men, how many are left-handed?
(b) Out of 100 left-handed adults, how many are men?
Please estimate:
(a) The probability that a woman is a school teacher.
(b) The probability that a school teacher is a woman. (p. 66)
32

Student performance on the conditional frequency item was higher than on the probability item.
Watson and Moritz claimed that the grammatical structure of the items affected performance
with the statement out of in the frequency item leading students to the relevant conditions. This
claim was in agreement with the findings by Pollatsek et al. (1987) who investigated college
students’ reasoning about conditional probability and found that they confused a conditional
event and its inverse.
Relative to research on biases associated with conditional probability, the work of Falk
(1979, 1989; see Batanero and Sanchez, 2005) is of significance. In particular, Falk
demonstrated the fallacy of the time axis using the responses of 88 university students to the
following problem:
An urn contains two white balls and two red balls. We pick up two balls at random, one
after the other without replacement. (a) What is the probability that the second ball is red,
given that the first ball is also red? (b) What is the probability that the first ball is red,
given that the second ball is also red?
Although students performed well in part (a), they were confused with part (b) because “they
thought that an event couldn’t condition another event that occurs before it” (Batanero and
Sanchez, 2005, p. 251). These students confused conditional with causal thinking. In particular,
they claimed that the answer to part (b) was 0.5 because the second event does not affect the first
since the second ball had not been drawn at the time of drawing the first ball (Falk, 1979, 1989;
see Batanero and Sanchez, 2005).
Falk (1986, 1988) pointed out a number of reasons as to why students find situations that
involve conditional probabilities difficult. One of the reasons was related to the grammatical
structure of items as also indicated by Pollatsek et al. (1987) and Watson and Moritz (2002),
whereas other reasons included: i) students’ difficulty recognizing the conditioning event; and ii)
33

students’ confusion of conditionality versus causality leading to a confusion of an event and its
inverse. That is, people often confuse P(A|B) and P(B|A) which Falk (1986; see also Diaz and de
la Fuente, 2007) termed the fallacy of the transposed conditional.
Another study related to misconceptions about conditional probability was carried out by
Gras and Totohasina (1995; see also Batanero and Sanchez, 2005). Participants included
seventy-five 17 to 18-year-old secondary school students. Responses were indicative of three
misconceptions relative to conditional probability: i) the chronological conception in which
students view that in P(A|B), the conditioning event B should always precede event A; ii) the
causal conception in which students view that the conditioning event B is the cause and event A
is the consequence; and iii) the cardinal conception in which students view that

P( A | B) 

Card ( A  B)
. It should be noted that the cardinal conception is correct in the case of
Card ( B)

an equiprobable sample space. Gras and Totohasina suggest that the chronological and causal
conceptions have a cognitive origin whereas the cardinal conception is brought on through
instruction (see Batanero and Sanchez, 2005).
Research carried out by Diaz and de la Fuente (2007) with 414 freshmen in a collegelevel introductory statistics course revealed that receiving formal instruction on probability was
unrelated to some of the biases relative to conditional probability. The participants had studied
conditional probability in secondary school and received two weeks of formal instruction on
conditional probability prior to completing the study questionnaire. Based on the results, even
after formal instruction, students persistently confused independence with mutual exclusiveness
and used the chronological conception of independence. Moreover, 31% of the students confused
conditional with joint probability and 59% employed the confusion of the transposed conditional
34

(p. 136). Although some biases remained persistent, some others improved. For example, after
instruction participants were able to easily compute “without-replacement” conditional
probabilities.
2.1.2.4 Binomial Distribution
Several researchers carried out studies pertaining to people’s conceptualizations of the
binomial distribution. Stokes (1972; see Krause, 2001) taught the two components of the
binomial distribution formula separately with reference to a fictitious substance named
“quirnon”, without teaching anything else about probability. Students were asked to respond to a
problem which was then revised into a second version as follows:
“

N!
 The number of different sequences of alpha and beta particles
r!(n  r )!

p r  (1  p) n r  The probability of a sequence”
Version 1: “If a quantity of quirnon emits eight particles what is the probability that four
of them will be alpha particles and four of them will be beta particles?”
Version 2: “If a quantity of quirnon emits eight particles, what is the probability that any
one sequence out of the set of all sequences that contains four alpha particles and four
beta particles will occur?” (as cited in Krause, 2001, p. 31)
When the first version of the problem was used, most students were unable to provide a solution
whereas when the second version was used most of them solved it. Krause criticized the study
indicating that i) it provided “no useful information about the students”; ii) since nothing else,
apart from the particular procedure, was taught about probability, there was “no reason for Ss to
realize that in the first problem they should take into consideration all of the possible
permutations”; and iii) the use of the term “sequences” in the second version “influences the
interpretation of the problem by cueing the procedure for counting the number of different
permutations” and thus, Stokes “communicates the “deep structure” of the problem” to the
students whereas students should bring the “deep structure” to the problem themselves (p. 32).
35

In a later study, Fischbein (1975) used the following problem with 13, 15, and 17 yearolds: “A warehouse contains 90% ripe apples, and 10% green apples. If three apples are taken at
random, what is the probability that two of them will be ripe?” Although, as Fischbein indicated,
participants had the necessary knowledge to solve the problem (that is, they were aware of the
addition law and the multiplication law), only 60% provided a solution. The common response
2

2

9 1
9 1
was   
instead of 3    
leading Fischbein to conclude that participants lacked
 10  10
 10  10
an intuition of a compound event. Since Fischbein did not clearly state the subjects’ level of prior
knowledge and math experience, an alternative explanation for this result may be that students
had ‘translation’ issues and misinterpreted the problem as asking for the probability of a “single
outcome” (Konold, 1989; Krause, 2001). Fischbein simply stated that students possessed the
necessary knowledge to tackle the problem but “what he probably meant was that they knew the
procedures needed to calculate the correct answer” which does not necessarily mean that subjects
understood the problem and its relation to these procedures (Krause, 2001, p. 30).
2.2 Probability Experiments with Real Data
In the attempt to aid students in overcoming any probabilistic misconceptions and
difficulties they may have, mathematics and statistics educators recommend that instruction on
probability be carried out in a way that is different from traditional (lecture-based) instruction
(Franklin and Garfield, 2006; MAA, 1998; Shaughnessy, 1977). The NCTM (2000) indicates
that “[D]oing mathematics involves discovery” and that students “should learn to investigate
their conjectures using concrete materials” (p. 57). Given this statement and considering the
increased demand by employers for university graduates who are able to work cooperatively to
solve problems in their work environments (Mackisack, 1994), it is essential that college
36

students are provided with opportunities that help them develop their problem-solving skills as
well as their verbal and written capabilities of presenting the results of their solution methods. In
order to help students develop such skills, instruction in introductory statistics courses should
include the use of experiments and real data collected and analyzed by the students themselves.
The American Statistical Association supports the use of real data sets which are of
interest to students in statistics classrooms (Aliaga et al., 2005; Franklin & Garfield, 2006).
Among the benefits of using such data sets is that they allow students to appreciate the difference
between empirical and theoretical approaches to explaining and predicting phenomena under
conditions of uncertainty (Batanero & Sanchez, 2005). Also, working with real data, points
towards the need for various probability distributions (Batanero & Sanchez, 2005).
In a study carried out by Amit and Jan (2006) with high-achieving students in grades 6-9
in Israel, game tasks involving coin tossing were used in the acquisition of probability concepts.
Participants had no formal background in probability and no formal teaching intervention was
implemented during the study. Instead, students built their knowledge of probability through
interactions with one another as they worked in triads. As a result of actively participating in
games, the students invented their own probability terminology and understanding of probability
concepts. In particular, students gained insight that there is a difference between theoretical and
experimental probability and that a link exists between probability and sample size.
Research by Shaughnessy (1977) with college undergraduate students in an elementary
probability and statistics course at Michigan State University, revealed that these students had a
positive attitude towards the course when it was taught in a small-group, experimental activitybased manner. In addition, in comparison to two lecture-based sections of the course, the two
experimental sections were more successful at helping students overcome probability
37

misconceptions. Day-to-day observations by the researcher in one of the experimental sections
indicated that “college students can learn to discover some elementary probability models and
formulas while working on probability experiments in small groups” (p. 313-314). The author
developed and used nine activities in probability, combinatorics, game theory, expected value,
and elementary statistics. The students in the experimental activity-based section worked on
these activities in groups of four or five students. The activities required students to perform
experiments, collect and analyze data, and reach conclusions based on their data. The approach
used in the experimental group was that of “problem-solving, model-building” with the
instructor circulating among the groups and using “the technique of ‘answering’ a question with
another question” so as to encourage students to think problems through for themselves (p. 299).
Shaughnessy (1981) recommends that an activity-based, experimental approach should be used
for the study of introductory probability and statistics since, according to evidence he gathered
through his research work, “an initial formalistic approach to probability is unlikely to help
students overcome misconceptions” (p. 95). Moreover, he suggests that instruction should
emphasize the use of simulations as a model-building and problem-solving tool.
In general, Garfield (1995) suggests that students learn better when they are involved in
hands-on activities which may be completed in cooperative small groups allowing them to apply
what they learn in new situations. Research carried out by Mackisack (1994) with sophomore
undergraduate students majoring in mathematics points towards the benefits suggested by
Garfield. The students in the Mackisack study were enrolled in a course that included design of
experiments and which was taught by the researcher herself. One of the course requirements was
that students work in small groups on a class project that involved experimentation in the form of
collecting and analyzing data, and presenting results to the whole class. The outcomes of the
38

study indicated that the group project helped students learn concepts relating to design of
experiments that would have otherwise been difficult to learn through standard textbook
exercises. At the same time, being directly involved in data collection allowed students to be
better able to critique experimental designs and question data interpretations.
In his work, Steinbring (1984; 1991; see Jones and Thornton, 2005) examined probability
both from its empirical and theoretical forms and emphasized the connection between the two.
Due to this connection, he supported the “simultaneous and mutual development” of the two
views of probability (p. 79). According to Steinbring, probability instruction should be carried
out as follows:
… learning begins with personal judgments about a random situation; comparisons are
made between the empirical situation and conjectured theoretical models, and finally
these comparisons lead to generalizations and more precise characterizations of the
random situation (Jones and Thornton, 2005, p. 78-79).
2.3 Cooperative Learning
Research carried out during the last few decades points towards the benefits of
instructional methods that use cooperation among multiple students along with student
engagement in explaining and reasoning with concepts (Aquilonius, 2005; Forman, 1996;
Franklin & Garfield, 2006; Zieffler et al., 2008). This section provides a review of i) the
characteristics of cooperative learning as an instructional method; ii) recommendations proposed
by professional organizations and researchers that relate to the use of cooperative learning in the
classroom; and iii) research outcomes (such as the effects on student attitudes and achievement)
of studies carried out in which cooperative learning was used.
As Dewey (1969) stated, the fact is that “all human experience is ultimately social: that it
involves contact and communication” (p. 38). An experience is based on the “transaction” taking
39

place between an individual and his environment; that is, the interactions one may have with
other people as they discuss an idea or with the materials used to perform an experiment (p. 4344). In the area of psychology, Vygotsky put forward the view of learning as a social process and
emphasized the role of dialogue in instruction. “The mere exposure of students to new materials
through oral lectures neither allows for adult guidance nor for collaboration with peers” (Cole,
John-Steiner, Scribner and Souberman, 1978, p. 131).
An instructional approach which is based on Dewey’s and Vygotsky’s ideas of educative
experience as a social process is that of cooperative learning. In such an approach, the instructor
is no longer the “boss or dictator” but the “leader” of group activities (Dewey, 1969, p. 59). It
should be noted though that assessing student learning under cooperative learning conditions is
more challenging (Smith, Douglas and Cox, 2009). To this end, it is recommended that students’
group work and group learning be encouraged and supported but that individual learning and
performance be assessed (Johnson and Johnson, 2004).
Researchers suggest that when designing tasks which involve the use of small groups the
following should be considered: i) the task should challenge all group members but the group, as
a whole, should be able to attend to it and complete it; ii) authority should be turned over to the
groups; iii) students should be assigned roles but these roles should not be static; and iv) groups
should consist of three to four students with different skills (Zawojeski, Lesh and English, 2003;
see also Kahveci and Imamoglu, 2007).
The importance of students communicating ideas with their classmates through tasks that
require cooperative group work is emphasized by the NCTM (1991, 2000). Students should be
provided with opportunities to work with classmates, to make and test conjectures and to listen to
explanations given by other students (NCTM, 2000). In this manner, they can learn to reason
40

through class discussions with fellow students; this may allow them to compare their ideas with
those of classmates and help them to modify or strengthen their own reasoning and clarify their
understanding thus developing their critical thinking skills (Garfield, 1993, 1995; Giraud, 1997;
NCTM, 2000). Since one of the content domains discussed by the NCTM is that of data analysis
and probability, opportunities for communication in cooperative groups should be provided to
students during instruction on probability.
In a study carried out by Leikin and Zaslavsky (1997) with four low-level ninth-grade
mathematics classes, the findings showed that students in the small-group cooperative learning
sections tended to be more active and remained more on-task while also having more
opportunities to receive help, than students who received instruction in a conventional way
(whole-class setting). In most cases that students in the cooperative learning sections asked for
help, this was provided in the form of explanations. In this manner, the cooperative learning
method allowed low-level students to engage in constructing explanations of the principles
needed for the solution of problems. Furthermore, the results showed that through the exchange
of knowledge, low-level students may develop their mathematical communications skills. In
support of this, explainers in Dansereau’s study (1988) learned more than students who received
explanations who, in turn, learned more than those working individually.
Relative to these results, a study by Webb, Tropper, and Fall (1995) with 7

th

graders

working in small groups on units on operations with decimal numbers and fractions, revealed
that the level of help received by a student from another student in the small group was a
significant predictor of constructive behavior (e.g. reworking the problem after receiving help)
which in turn was a strong predictor of post-test achievement. In general, student interactions and
the promotion of mathematical communication are essential in achieving high quality
41

mathematical learning (Bishop, 1985; Brown and Campione, 1986; see Leikin and Zaslavsky,
1997).
Results from a study by Dees (1991) carried out in a college remedial mathematics course
that included high school algebra and geometry indicated that cooperative learning helped
students increase their problem-solving skills. The course consisted of a large lecture section and
four laboratory sections to which students were randomly assigned. Two of the laboratory
sections served as the treatment group and the other two as the control group. In the treatment
group, students were organized with respect to cooperation whereas in the control group
cooperation was not actively encouraged (nor discouraged). Students who used cooperative
learning demonstrated significantly better results than students in the control group when it came
to solving word problems in algebra and writing proofs in geometry.
Chance and Garfield (2002) found that instructional methods that promote active
learning, such as cooperative learning, aid students in developing statistical reasoning.
Additionally, student misconceptions in probability may be overcome through working in small
groups (Gnanadesikan, Scheaffer, Watkins and Witmer, 1997; see also Pfaff and Weinberg,
2009; Shaughnessy, 1977, 1981; Webb, Tropper & Fall, 1995) on activities that confront such
misconceptions (Sáenz, 1998). Proponents of the use of small groups indicate that students may
learn better from their peers rather than from the instructor and that they may let go of a
misconception if they hear a convincing argument from a peer (Shaughnessy, 1978; see
Khazanov, 2005; Garfield 1995).
Another reason for advocating cooperative group work is that student achievement may
be improved through cognitively demanding tasks set by teachers (Giraud, 1997; Hiebert et al.,
2005; Potthast, 1999) which lead to multi-person communication among students and
42

simultaneously help maintain students’ interest (Silver, Mesa, Morris, Star, & Benken, 2009).
Dees (1991) found that students in a college remedial mathematics course performed the same
under the cooperative learning and traditional methods of instruction on algebra tasks that did not
involve complex thinking. However, students using the cooperative learning approach performed
better on measures that tested higher cognitive skills. A meta-analysis of research carried out by
Webb (1991) on the relationship between verbal interactions and learning in small groups in
mathematics classrooms indicated that giving content-related elaborate explanations to
teammates had a positive effect on achievement. In 50 out of 64 studies reviewed by Slavin
(1995) where the use of cooperative learning in which group rewards were provided, significant
positive effects on achievement were found. Meta-analyses which compared students’
achievement indicated that “the average student taught by cooperative learning performs better
than the average student taught with competitive and individualistic methods” (Johnson &
Johnson, 1989; as cited in Potthast, 1999; Johnson, Maruyama, Johnson, Nelson, & Skon, 1981).
Furthermore, the results of a study by Giraud (1997) support the hypothesis that “cooperative
learning … is especially beneficial for those least prepared for statistics” (paragraph 36).
The relation between cooperative learning and students’ attitudes has also been studied.
Results from research carried out by Zainun (2001) and Mazlan (2002) indicate that when
cooperative learning was used students held positive attitudes towards mathematics (see Effandi
and Zanaton, 2007). A meta-analysis of studies relating to small-group learning revealed positive
effects on achievement but also on persistence and attitudes among undergraduates in
mathematics classrooms (Springer, Stanne, & Donovan, 1999).
Through teaching an undergraduate introductory statistics course, Gunawardena (1998)
noticed that traditional instruction (lecturing) was not well suited for the students taking the
43

particular course at the particular research site. As a result, instead of using a traditional lecturebased instructional format, cooperative learning was used in the course during the subsequent
semester. In this study, the first half of each class period was devoted to the introduction of new
concepts whereas students worked in groups of three during the second half of the period.
Student groups performed such activities as reading the text, understanding the concepts
introduced at the beginning of class and solving problems. Students were also expected to work
in their groups outside of class to complete homework problems. The results revealed that the
new instructional method i) improved attendance; ii) increased the willingness of students to
participate in class discussions; iii) helped students learn statistics with less anxiety; and iv)
increased student-teacher interactions and student usage of office hours.
In another study carried out by Knypstra (2009) in an undergraduate econometrics
course, students were placed in small groups and were encouraged to work actively on specific
tasks. Cooperative learning and peer instruction were implemented with students explaining
theory and problem solutions to other groups. Course evaluations revealed that students felt that
they were more involved and more satisfied in this instructional format as compared to the
traditional lecture-based format.
In addition to the above, a study by Keeler and Steinhorst (1995) revealed that when
cooperative learning was used, more students completed the introductory statistics course.
Research in STEM education suggests that an essential element of success in college is positive
peer relationships whereas isolation is one of the best predictors for failure (Smith, Douglas, and
Cox, 2009). Among the major reasons for students dropping out of college is the inability to i)
form a network of classmates with whom they may work together, solve problems and discuss
the material taught and ii) be involved in the classroom (Tinto, 1993). Overall, Shaughnessy and
44

Bergman (1993; see Benko, 2006) suggest that instruction on stochastics would be improved
with the use of small groups in which students are actively involved in the exploration of
probability concepts.
2.4 Understanding
The term understanding is widely used in the mathematics education community. In
statistics education though, the term probabilistic understanding has not been explicitly defined.
Instead, researchers have established definitions of probabilistic reasoning and probabilistic
thinking both of which entail the idea of understanding. Since this study involves students’
understanding of probability, this section presents a review of the literature on mathematical
understanding, probabilistic reasoning and probabilistic thinking along with available models for
examining students’ probabilistic reasoning. Furthermore, the framework used in this study is
provided.
The literature includes various forms of understanding such as instrumental, relational,
conceptual, procedural, formal, algorithmic and intuitive. Some researchers advocate a specific
type of understanding (e.g. Skemp, 1978a, 1978b) whereas others claim that “it is the
combination and integration of the different types that is empowering” (Kvatinsky and Even,
2002, p. 4). This is especially true for the domain of probability where intuition may be
misleading and so, using other types of understanding “could serve as control” (Kvatinsky and
Even, 2002, p. 5).
2.4.1 Mathematical Understanding
According to Dewey (1933) “to understand is to grasp meaning” (p. 132) and this
meaning may be acquired by seeing something “in its relations to other things” (p. 137). In his
seminal work Skemp (1978a, 1978b) distinguished between three types of understanding in
45

learning mathematics: instrumental, relational and logical. Instrumental understanding involves
“rules without reasons” (Skemp, 1978b, p. 9); that is, it reflects students’ ability to use a rule
when solving a problem without knowing why the rule works. Relational understanding though
entails knowing what to do as well as why the rule works; that is, it entails an understanding
between the task at hand and the mathematical content, and requires an ability to explain why the
approach works. The third type of understanding, i.e. logical understanding, goes beyond
instrumental and relational understanding. It involves both knowledge of what to do and why but
in addition, it includes the ability to express the solution of the task correctly in written or
symbolic form; that is, it entails knowledge of the conventions of school mathematics (Skemp,
1978a, 1978b).
In his earlier work Skemp (1971) had stated that “to understand something means to
assimilate it into an appropriate schema” (as cited in Resnik and Ford, 1981, p. 167) where a
‘schema’ is “a conceptual structure” (Even and Tirosh, 2008, p. 206). Instrumental mathematics
involves the formation of short-term schemas through which students may get to the correct
answer more quickly. This type of mathematics makes use of an increasing number of rules,
which may be regarded as “degenerate schemas”, and which may be used successively in order
to get from one step to the next when working on a problem (Resnik and Ford, 1981, p. 168).
However, students learning instrumentally are not aware of the relationship between successive
steps. On the other hand, schemas formed by relational mathematics are more adaptable to new
situations which require conceptual connections (Resnik and Ford, 1981, p. 169).
Skemp advocated the use of relational mathematics and opposed to instrumental
mathematics (Even and Tirosh, 2008). However, several mathematics education researchers
raised concerns regarding the instrumental-relational dichotomy. For example, Resnik and Ford
46

(1981) argued that “automaticity of response”, which can be obtained by memorizing certain
procedures, can be beneficial because it frees up space in the working memory which can be
used for more complex procedures. Moreover, Hiebert and his colleagues (Hiebert and
Carpenter, 1992; Hiebert and Lefevre, 1986) as well as Star (2005) suggest that both procedural
and conceptual knowledge are important components of understanding.
The widespread use of conceptual knowledge and procedural knowledge can be credited
to the influential work of Hiebert and Lefevre (1986). In their book, conceptual knowledge is
defined as “knowledge that is rich in relationships … a connected web of knowledge, a network
in which the linking relationships are as prominent as the discrete pieces of information” (p. 3-4).
Procedural knowledge includes i) knowledge of the symbols and syntax and ii) knowledge of the
rules or procedures for manipulating symbols to arrive at a solution to a problem. Silver (1986)
argues that educators need to consider the relationships among these two types of knowledge and
not the distinctions between them because “distinctions are static, yet when one uses knowledge
to perform a non-trivial task, the knowledge is used dynamically” (Kaplan, 2006, p. 19).
Although the definitions provided by Hiebert and Lefevre (1986) for conceptual and
procedural knowledge have been influential, their use is problematic since they “suffer from a
entanglement of knowledge type and knowledge quality” (De Jong & Ferguson-Hessler, 1996;
Star, 2000; see Star, 2005, p. 408). As Star (2005) indicates,
The term conceptual knowledge has come to encompass not only what is known
(knowledge of concepts) but also one way that concepts can be known (e.g. deeply and
with rich connections). Similarly, the term procedural knowledge indicates not only what
is known (knowledge of procedures) but also one way that procedures (algorithms) can
be known (e.g. superficially and without rich connections) (p. 408).
Note that in his reference to procedures, Star specifies that these are algorithms. In his work, he
points out that there are various kinds of procedures and the type of connections associated with
47

each varies (see also Anderson, 1982). Algorithms constitute a set of steps that when followed in
a particular order they lead to the solution to a problem. “Algorithms are apparently what Hiebert
and Lefevre had in mind when they crafted their definition of procedural knowledge” (Star,
2005, p. 407). In such a case, Star agrees with Hiebert and Lefevre that (algorithmic) knowledge
is superficial and not rich in connections.
However, other procedures are heuristics which can be “tremendously powerful assets in
problem solving” (Star, 2005, p. 407). In particular, the use of heuristics requires that one makes
a choice and “wise choices can indicate quite sophisticated and deep knowledge” (p. 407). In the
case of probability, a substantial amount of research has been carried out relative to people’s use
of heuristics which has been reviewed earlier in this chapter (see section 2.1.1). Hiebert and
Lefevre’s definition of procedural knowledge though did not account for heuristics. Star argues
that if type and quality (two “independent” characteristics of knowledge) are separated then this
“allows for the reconceptualization of procedural knowledge as potentially deep” (p. 408). In his
words,
Deep procedural knowledge would be knowledge of procedures that is associated with
comprehension, flexibility, and critical judgment and that is distinct from (but possibly
related to) knowledge of concepts. (p. 408)
Recognizing that no single term captures entirely all that is entailed in mathematical
knowledge and understanding, the National Research Council (2001) referred to mathematical
proficiency which includes five “interwoven and interdependent” strands: conceptual
understanding, procedural fluency, strategic competence, adaptive reasoning and productive
disposition (p. 116). Conceptual understanding means the “comprehension of mathematical
concepts, operations, and relations” (p. 116). Students with this type of understanding know
more than facts and methods. They are actually able to form connections between new ideas and
48

their prior knowledge which allow them to reconstruct knowledge when forgotten. An important
indicator of conceptual understanding is being able to use various representations for
mathematical situations. Procedural fluency involves “carrying out procedures flexibly,
accurately, efficiently, and appropriately” (p. 116). The study of algorithms as “general
procedures” allows students to see that classes of problems can be solved using the same
procedure, pointing towards the fact that mathematics is well-structured (p. 121). For this reason,
“some algorithms are important as concepts in their own right” since they provide a link between
procedural fluency and conceptual understanding (p. 121). The third strand, i.e. strategic
competence, entails the “ability to formulate, represent, and solve mathematical problems” which
is similar to what has been termed “problem solving” in mathematics education (p. 116). This
strand includes the ability to first understand a problem situation and its key features, and then
provide a numerical, symbolic, verbal or graphical representation. When students are faced with
non-routine problems for which they do not immediately know a correct solution method, they
are required to find a way to understand and solve the problem. Adaptive reasoning is the
“capacity for logical thought, reflection, explanation, and justification” while the last strand,
productive disposition, is the “habitual inclination to see mathematics as sensible, useful, and
worthwhile, coupled with a belief in diligence and one’s own efficacy” (p. 116).
2.4.2 Probabilistic Reasoning, Thinking and Understanding
Unlike mathematical understanding, probabilistic understanding has not been defined in
the literature. Researchers defined terms such as probabilistic reasoning and probabilistic
thinking which entail the idea of understanding. This section considers these definitions while
the next section provides an overview of available theoretical frameworks developed as aids in
describing students’ probabilistic reasoning and thinking.
49

Jolliffe (2005), in agreement with Watson (2005), defined probabilistic reasoning as
“understanding and being able to explain and justify probabilistic processes” (p. 326), while
Garfield (2003) defined reasoning about uncertainty as
Understanding and using ideas of randomness, chance, likelihood to make judgments
about uncertain events; knowing that not all outcomes are equally likely; knowing how to
determine the likelihood of different events using an appropriate method (such as a
probability tree diagram or a simulation using coins or a computer program) (p. 25).
The second term that has received attention in the literature is probabilistic thinking
which Jolliffe (2005), along with Langrall and Mooney (2005), defined as
the way people reason with the ideas of probability and make sense of probabilistic
information. … probabilistic thinking involves understanding how models are used to
simulate random phenomena, how data are produced to estimate probabilities, and how
symmetry and other properties of the situation enable the determination of probabilities.
It also involves being able to understand and use context when solving a problem… (p.
326)
Jolliffe (2005) emphasizes that both probabilistic thinking and reasoning involve
understanding. Notice that the definitions for probabilistic reasoning and thinking make frequent
references to the term ‘understanding’. However, the term has not been explicitly defined
specific to the domain of probability.
2.5 Use of Understanding In This Study
Since the research questions posed in this study refer to students’ understanding of
probability, a definition of understanding as used here needs to be provided. Dewey (1933)
indicated that “to understand is to grasp meaning” (p. 132) and this meaning may be acquired by
seeing something “in its relations to other things” (p. 137). In a similar manner, Thompson and
Saldanha (2003) defined understanding as “the thought which results from a person’s
interpreting signs, symbols, interchanges, or conversation - assigning meanings according to a
web of connections the person builds over time” (p. 99). However, Thompson and Saldanha
50

point out that these connections are constructed through interactions that the person has with
“with his or her own interpretations of settings and through interactions with other people as they
attempt to do the same” (p. 99). Savery and Duffy (1996) pointed out that “understanding is in
our interactions with the environment” and “cognitive conlict or puzzlement is the stimulus for
learning” (p. 136).
The definition of understanding used in this study combines i) the definitions of
understanding provided by Dewey (1933) and Thompson and Saldanha (2003) and ii) the
definitions of probabilistic reasoning expressed by Garfield (2003) and Jolliffe (2005).
Therefore, students’ understanding of probability in this study means to grasp the meaning of
probability concepts by viewing them in relation to other concepts and to be able to explain the
processes used in the solution of probability problems or in the completion of probability
activities as students interact with one another in small groups. Such an understanding includes,
as Garfield stated, knowledge that not all outcomes have an equal probability of occurring and
the ability to use an appropriate method to determine the probability of an event. In this study,
the extended meaning of understanding given by Thompson and Saldanha (2003) of
“assimilation to a scheme” is used; this meaning allows for correct as well as incorrect or
inappropriate understandings (i.e. misconceptions) people may have. Researchers indicated that
“even errors are often signs of intelligent, although partial understanding of basic concepts”
(Resnik and Ford, 1981, p. 196). Therefore, in describing the understanding one may have we
need to be “addressing two sides of the assimilation – what we see as the thing a person is
attempting to understand and the scheme of operations that constitutes the person’s actual
understanding” (Thompson and Saldanha, 2003, p. 99; see also Liu & Thompson, 2007). As a

51

result, two things needed to be specified for each item on the pre-test and post-test used in this
study with regards to probabilistic understanding:
iii)

What is the goal of the item? That is, what probability content or concept is the item
addressing and the student trying to understand? and

iv)

What has the student actually understood or what misconception does he/she hold relative
to the particular item?
In this dissertation achievement is measured quantitatively by considering students’

responses to a probability pre-test and post-test. Understanding is measured in two ways: i) a
distractor analysis of student responses to the multiple-choice items on the pre-test and post-test
and ii) a qualitative analysis of audio-taped conversations as students worked on activities or
problem sets in groups during class. A distractor analysis provides useful, but not sufficient,
information on students’ understanding since multiple-choice items do not provide access to a
responder’s reasoning and subsequently, to a responder’s explanation as to why he/she chose a
particular response. Without access to a responder’s reasoning, an incorrect response to a
multiple-choice item does not directly indicate the use of a particular misconception; it simply
indicates the “application” of a given misconception (J. P. Smith, personal communication,
December 14, 2011). Therefore, both a distractor analysis and a qualitative analysis of student
conversations were carried out in the attempt to gain insight on students’ understanding of
probability in this study.
2.6 Available Models
2.6.1 Shaughnessy (1992): Stochastic Understanding
Shaughnessy (1992) modeled people’s stochastic understanding developmentally; this
understanding is characterized by the following four types:
52

1. Non-statistical. Indicators: responses based on beliefs, deterministic models,
causality, or single outcome expectations; no attention to or awareness of chance or
random events.
2. Naïve-statistical. Indicators: use of judgmental heuristics, such as representativeness,
availability, anchoring, balancing; mostly experientially based and nonnormative
responses; some understanding of chance and random events.
3. Emergent-statistical. Indicators: ability to apply normative models to simple
problems; recognition that there is a difference between intuitive beliefs and a
mathematized model, perhaps some training in probability and statistics, beginning to
understand that there are multiple mathematical representations of chance, such as
classical and frequentist.
4. Pragmatic-statistical. Indicators: an in-depth understanding of mathematical models
of chance (i.e frequentist, classical, Bayesian); ability to compare and contrast various
models of chance, ability to select and apply a normative model when confronted
with choices under uncertainty; considerable training in stochastics; recognition of the
limitations of and assumptions of various models (p. 485)
2.6.2 Jones, Thornton, Langrall & Tarr (1999): Probabilistic Reasoning
Jones, Langrall, Thornton and Mogill (1997) provided a framework for elementary
students’ probabilistic thinking across four constructs: sample space, probability of an event,
probability comparisons, and conditional probability. Around the same time, Tarr and Jones
(1997) described a framework for assessing middle school students’ thinking about conditional
probability and independence across four levels of thinking: Level 1 Subjective; Level 2
Transitional; Level 3 Informal Quantitative; and Level 4 Numerical. Subsequently, a framework
was developed by Jones, Thornton, Langrall, and Tarr (1999) which included the constructs and
levels of reasoning mentioned above, but in addition, included the construct of experimental
probability. The framework was tested and validated through the middle grades. The framework
may be used “as a filter for analyzing and classifying students’ oral and written responses”
(Jones et al., 1999, p. 153). Students’ level of understanding of a construct is revealed through
their ability to demonstrate certain behaviors when dealing with situations that involve
conditions of uncertainty.
53

Table 2.1
Framework for Students’ Probabilistic Reasoning (Jones et al., 1999, p. 15)
Construct

Level 1
Subjective

Sample Space

-lists an
incomplete set
of outcomes for
a 1-stage
experiment

Experimental
Probability of
an Event

-regards data
from random
experiments as
irrelevant and
uses subjective
judgments to
determine the
most or least
likely event
-indicates little
or no awareness
of any
relationship
between
experimental
and theoretical
probabilities

Theoretical
Probability of
an Event

-predicts
most/least likely
event on the
basis of
subjective
judgments
-recognizes
certain and
impossible
events

Level 2
Transitional

Level 3
Informal
Quantitative
-lists a complete set -consistently lists
of outcomes for a
the outcomes of a
1-stage experiment 2-stage
and sometimes for
experiment using
a 2-stage
a partially
experiment
generative
strategy
-puts too much
-begins to
faith in small
recognize that
samples of
more extensive
experimental data
sampling is
when determining
needed for
the most or least
determining the
likely event;
event that is most
believes that any
or-least likely
sample should be
-recognizes when
representative of
a sample of trials
the parent
produces an
population
experimental
-may revert to
probability that is
subjective
markedly
judgments when
different from the
experimental data
theoretical
conflict with
probability
preconceived
notions

-predicts most/least
likely event on the
basis of
quantitative
judgments but may
revert to subjective
judgments

54

-predicts
most/least likely
events on the
basis of
quantitative
judgments
-uses numbers
informally to
compare
probabilities

Level 4
Numerical
-adopts & applies
a generative
strategy that
enables a complete
listing of the
outcomes for 2and 3-stage cases
-collects
appropriate data to
determine a
numerical value
for the
experimental
probability
-recognizes that
the experimental
probability
determined from a
large sample of
trials approximates
the theoretical
probability
-can identify
situations in which
the probability of
an event can be
determined only
experimentally
-predicts
most/least likely
events for 1-and
simple 2-stage
experiments
-assigns numerical
probability to an
event (either a real
probability or a
form of odds)

Table 2.1 (cont’d)
Probability
Comparisons

Conditional
Probability

Independence

-uses subjective
judgments to
compare the
probabilities of
an event in two
different sample
spaces
-cannot
distinguish
“fair”
probability
situations from
“unfair”
ones
-following one
trial of a 1-stage
experiment, does
not always give
a complete
listing of
possible
outcomes for the
second trial
-uses subjective
reasoning in
interpreting with
and without
replacement
situations

-makes probability
comparisons on the
basis of
quantitative
judgments-not
always correctly
-begins to
distinguish “fair”
probability
situations from
“unfair” ones

-uses valid
quantitative
reasoning to
explain
comparisons and
invents own way
of expressing the
probabilities
-uses quantitative
reasoning to
distinguish “fair”
and “unfair”
probability
situations
-recognizes that the -recognizes that
probabilities of
the probability of
some events change all events changes
in a without
in a without
replacement
replacement
situation; however, situation
recognition is
-can quantify
incomplete and is
changing
usually restricted to probabilities in a
events that have
without
previously occurred replacement
situation

-has a
predisposition to
consider that
consecutive
events are
always related
-has a pervasive
belief that one
can control the
outcome of an
experiment

-begins to
recognize that
consecutive events
may be related or
unrelated
-uses the
distribution of
outcomes from
previous trials to
predict the next
outcome
(representativeness)
55

-can differentiate
independent and
dependent events
in with and
without
replacement
situations
-may revert to
strategies based
on
representativeness

-assigns numerical
probability and
makes a valid
comparison

-assigns numerical
probabilities in
with replacement
and without
replacement
situations
-uses numerical
reasoning to
compare the
probability of
events before and
after each trial in
with replacement
and without
replacement
situations
-uses numerical
probabilities to
distinguish
independent and
dependent events

A very similar framework to the one developed by Jones, Thornton, Langrall, and Tarr
(1999) was constructed by Polaki, Lefoka and Jones (2000). In this latter framework, the
researchers included probability of an event as a single construct instead of separating
experimental probability of an event and theoretical probability of an event like Jones et al.
(1999) did.
2.7 Theoretical Framework For This Study
The aforementioned models provide a set of key constructs of probabilistic reasoning
demonstrated in terms of students’ behaviors or learning goals that can be used to make sense of
students’ understanding of probability. Note that Shaughnessy’s (1992) model described
people’s understanding of stochastics in general in a developmental manner, whereas Jones et al.
(1999) focused on students’ reasoning along specific probability constructs.
The study presented here aimed to examine students’ understanding of experimental and
theoretical probability. Participants completed a probability pre-test and post-test along with a set
of in-class group activities (treatment group) or problem sets (control group). The activities and
problem sets focused on the following probability constructs: Law of Large Numbers,
experimental probability, theoretical probability, conditional probability, independence, discrete
probability distributions and the Binomial distribution. Since the data included both oral and
written student responses, a framework was needed to qualitatively analyze these responses.
Given that the instruments involved items on the constructs mentioned above, a framework was
needed that focused on these constructs and examined students’ level of understanding of them.
Therefore, the framework developed by Jones et al. (1999) was used in this study. The way in
which the framework was used to qualitatively analyze the data along with the results of the
qualitative analysis are presented in Chapter 6.
56

2.8 Summary
During the last few decades, probability has been gaining importance as an area that
students need to have experience with in order to be well-informed citizens. However, the
literature suggests that “there is no simple story about how students reason about chance”
(Konold et al., 1993, p. 413) making probability instruction a challenging task. Research has
revealed that both children and adults frequently hold intuitions about probability that come at
odds with theory. These problematic intuitions may explain why probability seems to be difficult
to learn. “One root of the trouble with probability is lack of experience with the long-term
regularity that the mathematics purports to describe” (Moore, 1997, p. 3). According to Konold
et al. (1993) one of the major reasons that probability is difficult to teach is that students do not
carry one but a variety of preconceptions regarding probability which they bring to the
classroom. Moreover, “[P]robability does not consist of mere technical information” and a set of
procedures but instead, requires a unique way of thinking that is different from other
mathematical domains (Fischbein and Schnarch, 1997, p. 104).
Shaughnessy (1981) suggests that instruction uses activity-based learning and considers
both an experimental and theoretical approach to the study of probability while the ASA
advocates active learning in introductory statistics courses (Franklin and Garfield, 2006; Aliaga
et al., 2005). Active learning may be fostered through the use of activities that involve group
work, cooperative learning, and class discussions (GAISE, 2005; see Hall & Rowell, 2008).
Furthermore, Steinbring (1991) supports that “empirical and theoretical probability should be
developed concurrently in the classroom” (as cited in Mojica, 2006, p. 10).
This dissertation study considers the recommendations set forth by the ASA and statistics
educators and examines the effects of two instructional methods on students’ understanding of
57

experimental and theoretical probability. The literature review helped identify instructional
characteristics advocated by researchers in mathematics and statistics education - such as group
work, active learning through the use of experiments that generate real data, and cooperative
learning – which were adopted in this study. Moreover, students’ probabilistic difficulties,
misconceptions and use of heuristics specified in the literature were considered in the
development of the instruments used in this study as well as in the analysis of student responses
on the instruments.

58

CHAPTER 3
CYPRUS EDUCATIONAL SYSTEM
As the current dissertation takes place in a tertiary education institution in Cyprus, this
chapter aims to provide information on the educational system of the country including the
intended mathematics curriculum. Consideration is given to the inclusion of probability both at
the secondary and tertiary levels since the participants in this dissertation study were beginning
college students in an introductory statistics course. Also, this chapter provides an overview of
the results of students in Cyprus on international studies specific to the domain of probability and
the consequences of these results on the Cyprus educational system. Moreover, a section is
included which considers the current educational reform efforts taking place in Cyprus and their
effect on the teaching of probability.
3.1 Control of the Educational System
Since 1960, the island of Cyprus has been an independent republic with a democratic
governmental system in which executive power is exercised by the president and legislative
authority is exercised by a house of representatives (Papanastasiou, 1997). The educational
system is highly centralized and controlled by the Ministry of Education and Culture which was
founded in 1965. Education in Cyprus is compulsory up to and including grade 9. Public as well
as private elementary, secondary and tertiary education institutions operate on the island and are
all liable to supervision by the Ministry of Education and Culture. Public education at all levels
is free whereas students attending private education institutions pay tuition. Entrance into most
private secondary schools is based on competitive exams in the Greek language and in
mathematics.

59

Until 1980 the influence of the Greek educational system on the Cyprus educational
system was “direct and unquestionable” with identical curriculum materials being used in
elementary and secondary schools in both countries (Lordou-Kaspari, 2003, p. 271). In 1980 “a
new system of specializations was introduced in the Cyprus Lyceum” (high school) (p. 271).
This meant that new curriculum materials were needed for grades 10-12. With this in mind, the
Ministry of Education and Culture established the Curriculum Development Service which in
turn developed the new curricula and textbooks needed for high school. In this respect, since
1980, the Cyprus educational system gained some degree of autonomy “although the links
between the Cyprus and the Greek education are still strong” (p. 271).
Currently, the Ministry of Education and Culture of Cyprus controls the curriculum, the
textbooks and other resources needed to deliver public education. Specifically, the textbooks
used in public elementary and lower secondary schools are either locally produced by the
curriculum development unit or donated by the Greek government (Pashiardis, 2007). The
curriculum in public schools is the same for all students up to the end of grade 9. Beyond grade 9
students may select some of the school subjects they take based on their interests.
In public educational institutions the teaching staff is appointed and promoted by the
Education Service Committee which operates within the Ministry of Education and Culture.
Private schools are operated and managed by individuals or bodies, yet they are supervised by
the Ministry of Education and Culture. Inspectors from the Ministry of Education and Culture
visit public as well as private educational institutions at all levels a few (unannounced) times
during the school/academic year (Pashiardis, 2007). The role of these inspectors is to attend class
sessions, observe whether students attend class and have acquired the course textbook, and
assess teacher performance. At the tertiary level, colleges and universities determine course
60

requirements for students with consultation with the Ministry of Education and Culture. All
private tertiary education institutions must register with the Ministry. If such an institution would
like to have its programs accredited then, in addition to registering with the Ministry, the
institution has to undergo an accreditation process.
Currently, higher education in Cyprus is provided by public and private tertiary education
institutions as follows: i) public or private universities; ii) non-university public higher-level
education institutions or iii) non-university private higher-level education institutions (colleges).
There are currently three public national universities (the University of Cyprus, the Cyprus
University of Technology and the Open University of Cyprus) and five private universities
(European University of Cyprus, Frederick University, Neapolis University, University of
Central Lancashire-Cyprus, and the University of Nicosia) in Cyprus. In addition, there are six
public colleges: the Higher Technical Institute offering courses in electrical, mechanical and civil
engineering; the Higher Hotel Institute; the Mediterranean Institute of Management; the Cyprus
Forestry College; the School of Nursing; and the Cyprus Police Academy (Lordou-Kaspari,
2003). Moreover, there are 24 private colleges offering diplomas or four-year university-level
degrees each specializing in certain areas of study including art, beauty therapy, accountancy,
banking, business administration, music, secretarial studies, technology, and hotel and catering.
It should also be noted that due to the geographic position of Cyprus, the island has become a
center of international business and shipping. As a result, most local higher-education institutions
offer courses in such areas as business administration, management, marketing, finance, and
accounting, attracting both home and overseas students.

61

3.2 Language of Instruction in Schools in Cyprus
The constitution of Cyprus identifies Modern Greek and Turkish as the two official
languages of the island. A Greek-Cypriot dialect exists which is unique to the island and it is
used in everyday spoken conversations by the, approximately, 700,000 Greek-Cypriots on the
island (Papapavlou, 2001). This dialect is closely related to Modern Greek but contains
influences from various languages including Latin, English and Turkish.
The economy of Cyprus depends highly on tourism, especially during the summer
months, and from the UK. Due to the geographic position of Cyprus at the crossroads of the
Middle East, Africa and Europe, several offshore companies operate on the island and employ
many Cypriots at their offices. Given these and the entry of Cyprus to the European Union in
2004, English has become a language that is used to a great extent on the island. Most job
openings require that candidates have a good knowledge of the English language; that they are
able to converse and write in English (Pavlou, 2000; as cited in Ministry of Education and
Culture, Cyprus, 2004). English is widely used in all government departments, in courts, and in
the banking sector (Papapavlou, 2001). Moreover, English frequently appears on shop signs and
billboards whereas local newspapers and television shows make frequent use of English words.
In schools, the teaching of English begins in the elementary grades with students taking English
language classes for two periods per week in grades 4-6 (Ministry of Education and Culture,
Cyprus, 2004).
In public secondary schools, instruction is carried out in Modern Greek but English is a
second language for which students take compulsory classes in grades 7 through 10 (Ministry of
Education and Culture, Cyprus, 2004). In contrast, in most private secondary schools, instruction
is carried out in English and students take compulsory classes in the Modern Greek language. At
62

most tertiary education institutions the language of instruction is English (Ministry of Education
and Culture, Cyprus, 2004) which was also the case at the research site for this study.
3.3 Intended Curriculum
The intended curriculum for all subjects taught in public elementary and secondary
schools in Cyprus is determined by the Ministry of Education and Culture. In the case of private
elementary and secondary education, the individual institutions determine the intended
curriculum with approval from the Ministry of Education. Mathematics is compulsory for all
students up to and including grade 12. This is the case in both public and private elementary and
secondary education institutions.
Specific to public schools, all students take the same mathematics course in the lower
secondary grades (grades 7-9/gymnasium). Upon entering upper secondary school (grades 1012/lyceum) students must register for electives, in addition to the compulsory common core
courses. Mathematics is offered as a common core course in grades 10-12. All students take
common core mathematics in grade 10. However, in grades 11 and 12 they may fulfill their
mathematics requirement by taking ‘core mathematics’ or by taking ‘advanced mathematics’.
At the secondary level, the aim of the mathematics curriculum is to develop in students
the ability to think logically, to analyze situations presented to them in the form of problems, to
understand concepts and their properties, and to use the language of mathematics (Papanastasiou,
1997).
3.3.1 Probability in the Intended Secondary Curriculum
Probability first appeared in the Cyprus curriculum during the school year 1984-1985 in
the algebra course for grade 9 (Lordou-Kaspari, 2003). Currently, probability is included in the
secondary school curriculum as part of the core mathematics course and the ‘advanced
63

mathematics’ course in grade 12 (Ignatiou and Zotos, 2007a; 2007b). The textbook used in the
core mathematics course (consisting of 198 pages) and the textbook used in the ‘advanced
mathematics’ course (consisting of 255 pages) each include a chapter on probability. In both
textbooks, the chapter on probability is rather short with the core mathematics textbook devoting
only 16 pages (i.e. 8% of the textbook) to this domain and the ‘advanced mathematics’ textbook
devoting 29 pages (i.e. 11% of the textbook) to it.
An examination of these two textbooks reveals that the intended curriculum on
probability in public secondary schools in Cyprus includes the following: i) definition of
randomness; ii) definition and identification of sample space and experimental outcomes; iii)
definition of probability and of various events including their properties (i.e. probability is a
number between 0 and 1; probability of a certain event is 1 and of an impossible event is 0; the
sum of the probabilities of complementary events is 1; if events A and B are mutually exclusive
then P(A∪B) = P(A) + P(B)); iv) definition and computation of the probability of simple events;
v) definition and computation of the probability of the union and intersection of events including
pictorial representations using Venn diagrams; vi) computation of the probability of an event
which involves the use of combinations and vii) tree diagrams and their use in solving
probability problems. The content described above is included in both the core mathematics and
the ‘advanced mathematics’ textbooks for grade 12 (Ignatiou & Zotos, 2007a, 2007b). In
addition, the textbook for ‘advanced mathematics’ includes a section on conditional probability
and a section on independence which are not part of the probability chapter in the grade 12 core
mathematics textbook.
With regards to probability problems, the core mathematics textbook for grade 12
includes thirteen problems in the probability chapter whereas the ‘advanced mathematics’
64

textbook includes a total of forty-eight problems in the corresponding chapter. In addition, each
of these two textbooks contains five problems on probability in the review exercises found at the
end of the book. Table 2.1 provides an overview of the content covered by these probability
problems.
Table 3.1
Probability Content In Grade 12 Mathematics Textbooks Used In Public Schools in Cyprus
Topic
Find the sample space
Compute the probability of an event
using probability properties
Compute the probability of simple
and joint events
Compute the probability of an event
using combinations
Compute conditional probabilities
Indicate whether two events are
independent; compute the
probability of independent events
Total Number of Problems

Number of Problems
Core Math Text
Advanced Math Text
3
3
3
11
6

13

6

7

0
0

12
7

18

53

3.3.2 Probability at the Tertiary Level
Probability content in tertiary-level courses in Cyprus is taught i) in a statistics course for
majors in the natural science, majors in the applied sciences or business, majors in the social
sciences, and education majors; ii) in a quantitative methods course for majors in business and
majors in education; and iii) as a research tool in programs in mathematics, statistics, education
and other disciplines (European University Cyprus, 2007; University of Cyprus, 2007; Frederick
University, 2008; University of Nicosia, 2008; Neapolis University, 2011; Cyprus University of
Technology, 2011; Open University of Cyprus, 2011; University of Central Lancashire-Cyprus
2012). Such courses are part of an undergraduate or graduate degree.
65

Note that among all tertiary education institutions in Cyprus, only the University of
Cyprus (which was the first university established on the island and began operating in 1992 as a
public university) and the newly established private University of Central Lancashire-Cyprus
offer undergraduate and graduate programs in mathematics and statistics. In other tertiary
education institutions in Cyprus, probability is only taught as part of service courses in
mathematics and statistics (including methods courses) for programs not in the natural sciences.
3.4 Performance of Secondary School Students on Probability on TIMSS
Cyprus has been a member of the International Association for the Evaluation of
Educational Achievement (IEA) since 1990 and has participated in the Third International
Mathematics and Science Study in 1995 and in almost all activities of Trends in International
Mathematics and Science Study (TIMSS) since then. Throughout the TIMSS studies, Cyprus has
been performing poorly on mathematics assessments. When the results of the 1995 Third
International Mathematics and Science Study were announced, the “exceedingly low
achievement of Cypriot students” sent a shock wave through the Cypriot society and the Ministry
of Education and Culture (Papanastasiou, 2002, p. 231). Furthermore, a comparison between the
1995 and 2007 TIMSS results as well as between the 1999 and 2007 TIMSS results indicated a
decrease in the mathematics achievement of 8

th

grade Cypriot participants (Mullis, Martin, &

Foy, 2008).
th

One of the content domains that TIMSS 8 grade participants are examined on is that of
data and chance. In the 2007 TIMSS the average scale score for 8

th

grade students in Cyprus on

data and chance was 464 (standard error 1.6) which was significantly lower than the TIMSS
th

scale average of 500 in this domain (Mullis, Martin, & Foy, 2008, p. 121). Cypriot 8 grade girls
66

averaged 474 (standard error 2.4) on data and chance which was significantly higher than the
th

average of 454 achieved by Cypriot 8 grade boys (standard error 2.5) in this content domain (p.
th

140). At the 8 grade in Cyprus, only 3% of class time is devoted to data and chance. Topics on
this domain taught up to and including grade 8 are considered to be only for the more able
th

students and as a result, only 3% of Cypriot 8 graders are taught such topics. That is, this 3% of
Cypriot 8

th

graders receive the 3% of instructional time devoted to data and chance. This is not

surprising since TIMSS 2007 and the intended curriculum discussion in section 3.3.1 revealed
that these topics are intended for grade 12 (Mullis, Martin, & Foy, 2008; Ministry of Education
and Culture & Pedagogical Institute, 2002).
3.5 Recent Developments in the Educational System of Cyprus
At the time of the announcement of the TIMSS 1995 results, the Ministry of Education
and Culture in Cyprus attempted to find excuses for the low achievement of students.
Certain circles in the Ministry of Education and Culture suggested that Cyprus ought to
have withdrawn from the international test for mathematics and science in good time
before it sank to the bottom, indicating the naïve attitude that a problem does not exist if
you do not see it (O Fileleftheros, 1996; see Papanastasiou, 2002, p. 231)
For some though the results were not surprising since the educational system was in need of
reform (Papanastasiou, 2002). In 1997 the International Institute for Education Planning
conducted a study of the educational system of Cyprus (Vrasidas and McIsaac, 2001). The
results of this study indicated that indeed the system was in need of reform in order for the
quality of public secondary education to be improved.
In view of these results, it is encouraging that since 2005 the government of Cyprus
initiated an Education Reform Program (The Ministry of Education and Culture, Republic of
67

Cyprus, 2008). A report by the Ministry of Education and Culture (2008) describes the plans and
areas in need of reform which include revisions in the national curriculum and the establishment
of a center for research. Among the key goals listed with regards to curriculum revisions, is to
help students develop into active citizens,
to enhance their critical thinking and research capabilities and to include a variety of
teaching methodologies and introduce flexibility in the school program, so that the
teacher may use the most appropriate approach for the particular class (The Ministry of
Education and Culture, Republic of Cyprus, 2008, p. 36)
Relative to teaching approaches, among the problems of the Cypriot educational system
listed in the UNESCO (United Nations Educational, Scientific, and Cultural Organization)
Report on Education in Cyprus, which was released after the TIMSS 1995 results, was that
teachers do not involve students in the learning process (Papastylianou, 1997; see also
Papanastasiou, 2002). In addition, based on the results of the study carried out in 1997 by the
International Institute for Education Planning the following should be integrated into the Cyprus
educational system: mixed ability classroom teaching, cooperative learning, and the use of
technology (Vrasidas and McIsaac, 2001).
In 2010, a detailed report comprising the reformed intended curriculum for each subject
area taught in grades K-12 was published by the Ministry of Education and Culture in
collaboration with the Pedagogical Institute and the Curriculum Development Service. These
reports, including one for mathematics, specify the content, procedures, applications and
experiences that students are expected to acquire while in school. Curriculum materials reflecting
the recommendations included in these reports were published in 2010. A small-scale pilot of
these materials was carried out during the school year 2010-2011 in grades K-9. Based on the

68

pilot results, the curriculum materials were to be revised by the end of the school year 2010-2011
and implemented in stages during the school year 2011-2012 in all grades.
The 475 page-long mathematics report of the intended curriculum is subdivided into five
domains: Number, Measurement, Geometry, Algebra and Statistics-Probability. The report
specifies that mathematics concepts should be taught in a way that advances students’ interest
and curiosity while simultaneously placing emphasis on problem solving. To this end, the
intended curriculum includes activities which require students to explore and discuss
mathematical ideas (Ministry of Education and Culture, Pedagogical Institute and Curriculum
Development Service, 2010). The general goals of mathematics education include that students
should: i) appreciate the value of mathematics and its use in all aspects of human activity; ii)
develop the ability to solve problems in multiple ways; and iii) develop the knowledge and skills
required in the workplace and for further studies in areas in which the use of mathematics is
necessary (p. 5). Moreover, the mathematics intended curriculum aims to provide opportunities
for students to: i) develop flexibility and creativity in the application of mathematics concepts in
problem situations; ii) solve problems cooperatively, express their ideas and respond to the ideas
of their classmates; iii) learn through the correct as well as incorrect responses given by them and
their classmates; and iv) develop their verbal, written and presentation skills so as to be able to
express and define mathematics concepts (p. 6). Relative to the domain of probability, Tables 3.2
- 3.4 provide the breakdown of skills that students in grades 10-12 are expected to acquire
(Ministry of Education and Culture, Pedagogical Institute and Curriculum Development Service,
2010, p. 440-466) (trans.):

69

Table 3.2
Probability in the Reformed Intended Curriculum for Grade 10 in Cyprus
Grade 10 – Probability – Student Skills
1. Represent events using Venn diagrams (union, intersection, difference,
complement)
2. State and apply the Kolmogorov axioms. Reason through them using suitable
examples and Venn diagrams. Apply the consequences of the Kolmogorov
axioms (( P( A), P( A  B), P( A  B) ) in solving problems.
3. Study random experiments involving two or more steps and write down the
sample space using tables and tree diagrams.
4. Understand and apply the Counting Principle in random experiments.
5. Convert tree diagrams into probability tree diagrams and compute the
probability of compound and independent events.
6. Compute permutations and combinations and apply these in the computation
of probabilities.

Table 3.3
Probability in the Reformed Intended Curriculum for Grade 11 in Cyprus
Grade 11 – Probability – Student Skills
1. Distinguish events as conditional and independent and compute their
probabilities.
2. Understand and apply the rule of Total Probability.
3. Understand the meaning of random variable.
4. Find the probability density function and the distribution function of a discrete
random variable.
5. Compute the mean and standard deviation of a discrete random variable.
6. Study the Binomial distribution.

Table 3.4
Probability in the Reformed Intended Curriculum for Grade 12 in Cyprus
Grade 12 – Probability – Student Skills
1. Understand the meaning of a probability density function and distribution
function of a continuous random variable.
2. Compute probabilities under specific intervals; find the mean and standard
deviation of a continuous random variable.
3. Know the properties of the Normal distribution. Solve problems by applying
the table of the standard normal distribution and the Central Limit Theorem.
70

3.6 Research Site for this Study
In view of the various statements made by the Ministry of Education and Culture of
Cyprus and the problem identified in the UNESCO report, as well as considering that “demands
for dealing with data in an information age continue to grow” (Franklin & Garfield, 2006, p.
363), it is important to study the effects of different teaching approaches in statistics classrooms.
A first step into such an endeavor – since there is a lack of research in this area specific to
Cyprus, especially at the secondary and post-secondary levels - was carried out through this
dissertation study.
The research site for this study was a specialized private business college operating since
1983 as a tertiary education institution in Cyprus with approval from the Cyprus Ministry of
Education. Instruction at this college is carried out in English. Most students attending this
college are native Cypriots who have graduated from a public high school and for whom English
is a second language.
At this college, probability is taught i) in an introductory business statistics course which
is compulsory for all freshmen and ii) in an Operations Management course required by junior
Business Computing majors and which involves applications of mathematics, statistics and
probability to the business world.
The study took place in the introductory statistics course at this college during spring
2010. Until that time, the course was being taught using a lecture format with complete reliance
placed on the course textbook. As mentioned in Chapter 1, the MAA (1998) and research in the
STEM fields promote the use of teaching techniques which involve less lecturing, increased
group work and student interactions in undergraduate classrooms. Moreover, it has been
recommended that statistics classrooms use a frequency-based approach to teaching probability
71

along with an active learning approach that includes the use of activities (Cobb, 2000; Watson,
2006). This study aimed to adopt these recommendations in the aforementioned introductory
statistics course. The methods used in accomplishing this are described in detail in the Chapter 4:
Methods.

72

CHAPTER 4
METHODS
4.1. Overview of the design
A mixed methods design was used to address the research questions for this study. The
design included treatment and control groups, each comprised of students in sections of an
introductory statistics class, and utilized a pre-test and post-test design. During the study students
in the treatment group worked in small groups on four in-class activities about experimental and
theoretical probability, and students in the control group worked in small groups on solutions to
four sets of probability problems. In each group, the conversations of specific students were
audio-recorded across the four occasions that they worked in groups. These data sources allowed
for both quantitative and qualitative analyses of data.
4.2. Research Site
4.2.1 College
The college at which the study took place is situated in Larnaca, one of the main cities in
Cyprus. It has been operating since 1983 as a tertiary education institution with approval from
the Cyprus Ministry of Education. The college is a specialized business school at which students
may pursue a four-year Bachelor’s degree in one of the following areas: Business
Administration, Business Computing, Accounting, or Banking. Alternatively, students may
pursue a two-year diploma in business administration or computing and information systems. In
addition, the college offers one-year certificates in law or business and information technology.
The 4-year Bachelor’s degree programs in Business Administration and Business Computing
have already been approved by the Cyprus Ministry of Education and most students attending the
college major in one of these two areas. During the academic year 2009-2010, all freshmen
73

pursued one of these two degrees. Approval for the rest of the programs offered at this college is
currently being sought.
At this college, the mathematics and statistics requirements for students in the Business
Administration program differ slightly from those for students majoring in Business Computing.
Table 4.1 below presents listings of the requirements for the two programs in the order in which
students take the courses.
Table 4.1
Mathematics Course Requirements For Students At Research Site
Business Administration
MAT 101 – Calculus
MAT 201 – Statistics I
MAT 202 – Statistics II
MAT 203 – Quantitative Methods

Business Computing
MAT 101 – Calculus
MAT 201 – Statistics I
MAT 210 – Discrete Mathematics
MAT 203 – Quantitative Methods
MGT 312 – Operations Management

Most of the students attending this college are native Cypriots. A majority of students
understands and speaks Modern Greek and most of them are using English as a second language.
The official language of instruction for all programs is English.
The academic year officially begins in the first week of October and finishes at the end of
May. During the entire month of September, freshmen take a foundations course in each of the
following subjects: mathematics, English, and accounting. At the end of September these
students take a placement exam in each of these subjects. The aim of these courses is to help
incoming students attain the basic skills in these subjects that will be required during their
academic studies. The material covered in the mathematics foundations course includes real
numbers and their properties, rules of exponents, simplifying algebraic expressions, and solving
linear and quadratic equations.
74

The college offers morning classes as well as evening classes. For example, students may
take a 4-credit course, such as MAT 201 – Statistics I, by enrolling in a morning section that
meets for four 50-minute periods a week between the hours of 8:15am-1:30pm or by enrolling in
an evening section of the course which meets once a week from 5:50-8:20pm. Evening sections
tend to be smaller than morning sections. Courses are carried out in classrooms that can hold up
to 30 students. All classrooms in which the introductory statistics course is accommodated are
equipped with an overhead projector as well as a LCD projector.
College students in Cyprus are required by the Ministry of Education to acquire and use a
textbook for every course they take. At the college where the study took place, all course
textbooks are written in English and are selected by the course instructor in consultation with
previous instructors of the same course, the department head and the course coordinator.
4.2.2. Course
The study took place in an introductory statistics course taught by the researcher in spring
2010 at the aforementioned college. This course (MAT 201 – Statistics I) is offered in the spring
semester of every academic year and it is compulsory for all freshmen attending the college.
Students taking the course are split into two morning sections (Sections A and B) and one
evening section (Section C). Since each morning section of the course meets for four 50-minute
sessions each week whereas the evening section meets for one 150-minute session each week,
students in the morning sections spend 50 minutes more per week in the course. These meeting
times are kept constant throughout the spring semester which lasts for 15 weeks (13 weeks of
teaching and 2 weeks of final exams). During the spring semester 2010, to control for the factor
of contact time for this study, students in the evening section were asked to attend four extra

75

class sessions from 5:50 – 8:20 pm. An arrangement was made between the instructor of MAT
201 and the students in Section C based on the students’ class and work schedules.
The introductory statistics course (MAT 201) covers the following topics:
i. Basic concepts of statistics (population, sample, parameter, statistic,

categorical /

numerical / discrete / continuous data);
ii. Tables and charts for presenting categorical or numerical data (bar charts, pie charts,
histograms, contingency tables, and scatter plots);
iii. Measures of central tendency and variation;
iv. Covariance and correlation;
v. Basic probability concepts (sample space, simple and joint events, Venn diagrams, union
and intersection of events, finding the probability of events);
vi. Conditional probability, independence, and Bayes’ Theorem;
vii. Counting rules;
viii. Discrete probability distributions (expected value and variance of discrete random
variables; binomial distribution);
ix. and the normal distribution.
The main course textbook for MAT 201 was a U.S. publication of Business Statistics: A
First Course (Levine, Krehbiel & Berenson, 2010). During this study, problems from this
textbook were selected by the researcher/instructor and were assigned as homework to the
students in all sections of the course.
4.2.3 Timeline and Procedures
During the summer of 2009, the instruments that were to be used in the study were
developed. For the purposes of this, the researcher carried out a thorough search in statistics
76

education and mathematics education journals, conference proceedings, statistics textbooks,
statistics education and mathematics education textbooks, instruments used in international
studies in mathematics education, and websites developed by mathematics and statistics
educators. Through this search it became evident that finding items and especially appropriate
activities specific to probability was a challenging task, since these were not commonly found
and many were not suitable for the purposes of this study. Information provided by a committee
member led to contacting the researchers involved in the ARTIST (Assessment Resource Tools
for Improving Statistical Thinking) project who in turn gave access to the researcher of this study
to the item bank as well as the instruments developed and tested by the ARTIST group. Once the
instruments for this study were developed, they were reviewed by committee members and by a
mathematics education graduate student at Michigan State University. Items were piloted in late
January 2010, and revised shortly thereafter.
nd

Instruction on probability formally began in week 7 (Monday, March 22 , 2010) and
st

continued until week 13 (Wednesday, May 21 , 2010) of the Spring 2010 semester. The pre-test
and background questionnaire were administered during week 7. In the treatment group, the
administration of the four activities on experimental and theoretical probability commenced
during week 7 and ended in week 13. On the same weeks that students in the treatment group
worked on the probability activities, students in the control group worked on sets of probability
problems during class. The post-test was embedded in the course final examination which was
th

administered at the end of week 15 and specifically on June 4 .
In order to motivate students to take the completion of the instruments seriously,
completing each instrument counted towards the student’s final course grade. Course policy at
77

the college where this study took place specifies that the final exam should count for 70% of the
course grade. The other 30% is allotted, according to the instructor’s judgment, to class
participation, a midterm, and a formal course assignment common to all sections of the course.
The formal course assignment for all sections of MAT 201 consisted of problems on statistics
only, counted for 8% of the course grade, and was due at the beginning of week 7, the same
week that students worked on Activity 1 and formal instruction on probability began. The
probability items on the post-test (embedded in the course final exam) were used to determine
individual mastery of the material covered on probability.
In all three sections of MAT 201, the course grade was split as follows:
Course Assignment on Statistics

8%

Class Participation

12%

Midterm exam

10%

Final Exam

70%

The percentage allotted to class participation (12%) was split as follows: In the case of the
treatment group (Section A and Section C), students received 2% for each of the four activities
on probability they participated in, for a total of 8%; in the case of the control group (Section B),
students received 2% for each of the four occasions they worked in groups to solve probability
problems from the textbook, for a total of 8%; students who completed the pre-test received 4%
to count towards their course grade.
During summer 2010, the researcher coded student responses to the background
questionnaire, the pre-test and the multiple-choice items on the post-test and entered these data in
SPSS. Quantitative data analysis of these data was carried out during summer 2010. With regards
to the free-response items on the post-test, rubrics were created during fall 2010. Student
78

responses to the free-response items were then coded and quantitatively analyzed using SPSS.
Data from the audio recorded group conversations were transcribed in December, 2011 and
January, 2012 and were then qualitatively analyzed.
4.2.4 Pilot
As previously mentioned, all instruments were piloted in January 2010. The pilot aimed
to reveal any possible misinterpretations caused by the instructions to the probability problems
on the instruments or the instructions to the probability activities and any obstacles faced by the
students due to specific vocabulary used in these problems and activities. The pilot also helped in
determining the length of time it would take to complete each activity, and how to handle
technical aspects of recording conversations of groups of students.
th

Piloting began on January 18 , 2010, once the study received IRB approval, and lasted for
two weeks of instructional time at the college. This time period marked the end of the fall
semester at the college (February 8

th

was the commencement of the spring semester).

Participants were 25 (out of 28) students taking MAT 202 Statistics II and four students taking
MAT 210 Discrete Mathematics that semester (i.e. Fall 2009). These students had received
instruction on the probability concepts covered on the instruments during the spring semester of
2009 in MAT 201 Statistics I. The researcher was not the instructor for MAT 202 but was the
instructor of MAT 210 during the pilot process. The completion of the instruments on the pilot
study did not count towards the course grade for any of the participants in the pilot. In order to
motivate participants in the pilot to work on the instruments, the researcher treated participants
from MAT 202 on two occasions to coffee and breakfast treats at the college cafeteria and gave
the course handbook for MAT 210 to the four students taking that course as a gift.
79

Once the study received IRB approval, the researcher visited the sections of MAT 202 and
MAT 210, talked to the students about the research study and asked for volunteers. Since it was
the end of the semester, and the mathematics instructor of MAT 202 had completed instruction
on the course material specified on the syllabus, he allowed the researcher to use the remaining
of the class sessions of MAT 202 prior to the end of the fall semester for the purposes of the
pilot. At the beginning of the first data collection meeting, the consent form (see Appendix A)
was read to the students, the study was explained once more, and participants were asked to sign
the consent form if they agreed to participate. The consent forms were placed in an envelope and
returned by one of the students in each course to the course coordinator. The envelope was given
to the researcher after semester course grades for MAT 210 and MAT 202 had been submitted.
All four students in MAT 210 and 25 of the 28 students in MAT 202 had volunteered to
participate.
The background questionnaire and pre-test were piloted during a 50-minute session. With
regards to the probability activities, two or three groups of volunteers worked on each activity
and the conversations of all groups were recorded. Audio recordings took place in regular
classrooms where the participants and the researcher met. This aimed to create conditions similar
to those that the researcher was to encounter in spring 2010 when recording group conversations
in the regular classroom. Such recordings during the pilot study helped the researcher identify
noise levels during group conversations and their effect on the quality of audio recording. This
indicated to the researcher how groups should be positioned in the classroom during spring 2010
to ensure better quality recording.
In addition to piloting the instruments with the students in MAT 202 and MAT 210, expert
reviewers were asked to provide feedback on the instruments during fall 2009. This included
80

expert reviewers who were members of the researchers’ dissertation committee. Based on this
feedback, in October 2009 the instruments underwent a first round of revisions with regards to
content and format. Relative to the pre-test and post-test, experts’ recommended revisions were
sorted from worst item to best item. These recommendations were then used to modify or delete
items.

In addition, in December 2009, a mathematics graduate student at Michigan State

University provided further feedback on the instruments. Upon the completion of the pilot at the
end of January 2010, the instruments underwent further revisions with regards to content, format,
and duration, based on the pilot data.
4.3. Participants
Prior to the start of the academic year 2009-2010, participants were split into three sections
(Sections A, B, and C) by the course coordinator in consultation with the academic board at the
college. During the academic year 2009-2010, the academic board decided to try a new approach
to placing freshmen in groups, based on their English language proficiency. As a result, Section
A (morning section) included the students of moderate/high English language proficiency
whereas Section B (morning section) included the students of low English language proficiency.
Based on the students’ course selection, all four of the students who majored in Business
Computing were placed in Section A, although two of these students were of low English
language proficiency. Section C (evening section) included students of all levels of English
language proficiency. Volunteers for this dissertation study came from all three sections of MAT
201 – Statistics I who had been split into three sections: Section A (morning) which consisted of
22 students, Section B (morning) which consisted of 19 students and Section C (evening) with 7
students. Out of these 48 students, 44 agreed to participate in the study by signing a consent
form: 20 from Section A, 17 from Section B, and 7 from Section C.
81

Participants were aged 18-27 years old with a mean age of 19.5. Twenty of them were
male (45.5%) and twenty-four were female (54.5%). Most of these students (42 out of 44) had
attended a public secondary school in Cyprus whereas one of them had attended a private
secondary school in Cyprus and one of them a secondary school abroad. Most of the participants
(40 out of 44) were native Cypriots with two from Russia, one from Georgia and one from
Moldova. All of them were able to communicate orally in Modern Greek, the native language in
Cyprus, and 41 of them were using Modern Greek as their first language and English as their
second language. Thirty six of the participants indicated that they had received instruction on
statistics and probability in high school whereas 7 of them indicated that they had not (1 student
did not respond). Forty three were freshmen whereas one of the participants was a sophomore.
This student (sophomore) had taken MAT 201 for the first time in spring 2009, with the same
instructor, but had not passed the course then. With regards to the various programs of studies,
32 of the 44 participants majored in Business Administration, 8 majored in Accounting, and 4
majored in Business Computing.
4.4. Instruction
The researcher was also the course instructor of MAT 201 in spring 2010. For the purposes
of this study, during the spring semester of 2010, one of the morning sections (Section A) and
the evening section (Section C) of MAT 201 – Statistics I were taught using an instructional
method that combined lectures and small-group cooperative learning sessions with students
performing experiments that generated real data and completing probability activities. These two
sections of MAT 201 formed the treatment group. The second morning section (Section B) acted
as the control group. In the control group, the researcher used an instructional method that

82

combined lectures with small-group cooperative learning sessions during which students worked
in class on solving probability problems from the course textbook.
Upon enrollment in the course, students did not know that different instructional methods
were to be used in the course sections. During the first session of MAT 201 in spring 2010, the
researcher/instructor talked to the students in the three sections of the introductory statistics
course about this dissertation study. The researcher indicated to the students that participation
was voluntary and explained the benefits as well as any risks regarding their participation.
Furthermore, the researcher pointed out to the students the purpose of the study, the necessary
IRB procedures and the need of signing a consent form in case they agreed to participate (see
Appendix A). According to the IRB, the researcher/instructor should not have access to the
signed consent forms until final course grades had been submitted. As a result, after talking to
the students about the study during the first class session of the course, the researcher/instructor
left the room and the signed consent forms were placed in an envelope and returned to the course
coordinator by one of the students in each section.
Meanwhile, the three sections of MAT 201 needed to be assigned as treatment or control.
In order for this assignment to be determined, the two morning sections which were of an almost
equal size (20 students in Section A and 17 students in Section B) were randomly assigned as
treatment or control by tossing a coin to determine what Section A would act as (heads for
treatment and tails for control). The evening section (Section C – 7 students) was assigned as a
treatment group.
4.4.1. Treatment Group
nd

During weeks 7 – 12 (March 22

th

– May 14 ) students in the treatment group worked in

small groups during class to complete four activities on probability which covered the following
83

topics: i) Law of Large Numbers; ii) Conditional probability and independence; iii) Discrete
probability distributions; and iv) the Binomial distribution. The activities aimed to bring about
the bidirectional relationship between experimental and theoretical probability and to prepare
students for the study of theoretical probability. Each activity required the students to perform a
probability experiment, to collect and analyze data and reach some conclusions. Students in the
treatment group received 2% for working on each of the four activities, for a total of 8% to count
towards their course grade. All activities appear in Appendix C.
The aim of Activity 1 (McConnell et al., 1998; see Appendix C) was to provide students
with a hands-on experience in viewing the connection between experimental and theoretical
probability, leading to the Law of Large Numbers. Two dice of different colors were given to
each group and students were asked to roll them 50 times. The group’s frequency and relative
frequency for each sum was recorded in a table. Students had already received instruction on
frequency, relative frequency and frequency histograms, so they were familiar with the terms
used in the activity. Once students carried out the experiment, they then examined the theoretical
probability of each sum of two dice by considering a pictorial representation of the sample space
that was provided in the activity. Next, the results of the 50 rolls of each group were recorded on
the class board and each group used these results to calculate the relative frequency for the entire
class. Students compared their group’s experimental results with the theoretical results. Also,
they compared the combined class results from all groups with the theoretical results, in order to
see that as the number of simulations increases the more the relative frequency approaches the
theoretical probability.
In Activity 2 (see Appendix C), students investigated the ideas of conditional probability
and independence. This was done through an exploration of the sum of three dice. Students were
84

asked to roll three dice of different colors simultaneously 50 times and record their results along
with the sum of the three dice in a table. Based on their group’s results, they were then asked to
compute the relative frequency of a simple event (i.e. one of the three dice comes up 3) and then
the relative frequency of the same event this time based on a pre-set condition (i.e. the relative
frequency of one of the dice comes up 3 given that the sum is 6). The students then explored the
probabilities of these two events through a theoretical approach by responding to a set of
questions which were ordered in such a way as to carefully guide them to the theoretical
conditional probabilities and the idea of independence.
Activity 3 (Khazanov, 2008; see Appendix C) allowed students to investigate discrete
probability distributions. The ideas presented here were adopted from a proposed activity
provided by Khazanov (2008). This activity involved a game of chance that made use of dice and
chips. Students had to place a set of 12 chips given to each of them along a number line, on
which numbers 1 through 15 were written, so as to increase their chances of having them
removed first – and thus win the game – based on the sum of two dice. A designated student
rolled the two dice, the sum of the numbers showing on the dice was calculated, and any student
in the group that had chips placed above the number represented by the sum removed one of
those chips. The winner was the student who had all of his/her chips removed first. Students
were asked to provide reasons for the way they each distributed the chips on the number line.
They were also asked to indicate the mistake one could make to eliminate their chance of
winning. In the last part of the activity, the students explored the discrete probability distribution
of the sum of two dice theoretically. In this last part, a table was provided to students listing the
36 possible outcomes when two dice are rolled. Students were asked to use this table to compute
the probability of each sum from 2 to 12 when a pair of dice is tossed. This procedure aimed to
85

help students understand why they may or may not have made poor choices when placing their
chips on the number line, and to realize that if they placed more chips towards the center of the
distribution they would have a higher chance of winning.
In the last activity (Shaughnessy et al., 2004), students explored the binomial distribution
first through the use of a table of random numbers and then theoretically. The researcher used
Excel to generate sets of random numbers from 1 to 10. A different set of such numbers was
given to each group. Students were instructed on how to use their corresponding table of random
values and were asked to use it to indicate data on 50 sets of three free-throw attempts for a
basketball player with a 70% free-throw average. On the basis of these 50 sets of simulated data,
students then computed the probability that the player will make at least two baskets in three free
throws. Next, students compared their results with those of one other group in class. In the
meantime, the instructor recorded the results of at least two baskets in 50 sets of three free-throw
attempts of all groups in class. These combined class results were used by the students to
compute the probability of making at least two baskets and compare their group’s answer to that
derived based on the class results.
During the four occasions of working on activities, students were placed in groups of 2 - 4
as determined by the instructor. An attempt was made to include in each group students of
varying mathematical ability in order to create opportunities for scaffolding (Vygotsky, 1978; as
cited in Giraud, 1997). Group members were assigned roles by the researcher/instructor so that
all members would have an opportunity to participate. These roles were rotated on every
occasion that students worked on an activity. The role of each student was indicated in written
form on the top of the first page of each activity handed to the group. Three roles were used in
the probability activities: performer of experiment (e.g. the person simulating rolls of a die as
86

required by the experiment), recorder of data and of group responses to activity questions, and
checker of data recordings and group responses to activity questions. In the case where a group
consisted of two students, one of the students was assigned a dual role: performer of experiment
and checker. In the case where a group consisted of 4 students, two of the students were assigned
the same role. At the beginning of each activity, the researcher briefly described the meaning of
each role to the whole class. All students in each group were encouraged to help the members of
their group and discuss their ideas on how to perform the experiment, how to record and analyze
the data and discuss what they thought were possible conclusions they could draw from their
data. The researcher/instructor circulated in the classroom as the students worked in their groups
and tried to monitor and encourage students to participate. Students were asked to submit a copy
of their group’s responses to the activity to the instructor at the end of class. Feedback was
provided to the students at the beginning of the next class session.
The class session following the completion of an activity began with providing feedback to
the students on their group work, reviewing the conclusions that could be drawn from the
activity, and looking at the bidirectional relationship between experimental and theoretical
probability as a whole class. These were completed during the first 15-20 minutes of the 50minute class session. The remaining class time was devoted to the study of probability
terminology and theoretical probability on the probability topic of that week using a lecture
format.
Following the lecture session on probability terminology and theory, the instructor solved a
few probability problems from the course textbook on the board relative to the probability topic
under study that week. The students were then assigned 2-3 problems from the course textbook

87

to work on individually as homework. These were corrected by the instructor and handed back to
the students.
4.4.2. Control Group
During weeks 7-12 (March 22

nd

th

– May 14 ) students in the control group worked in small

groups on four sets of probability problems assigned from the course textbook. The problems
covered the following topics: i) basic concepts of probability; ii) conditional probability and
independence; iii) discrete probability distributions; and iv) the Binomial distribution (See
Appendix C). Students in the control group received 2% for each set of problems they worked
on, for a total of 8% to count towards their course grade.
Similar to the treatment group, student groups in the control group were set up by the
instructor and consisted of 2-4 students. An attempt was made to include in each group students
of varying mathematical ability in order to create opportunities for scaffolding (Vygotsky, 1978;
as cited in Giraud, 1997). Group members were assigned roles by the instructor so that all
students would have an opportunity to participate. These roles were rotated on every occasion
that students worked in groups. A group member undertook one of the following three roles:
recorded the group’s solution to a problem, checked solutions, or asked questions to stimulate
group conversation. The first two correspond to the roles of recorder and checker used in the
treatment group. Instead of a student who would ask questions to stimulate group conversation,
the role of performer of the experiment was used in the treatment group. On the occasions that
students worked in groups, the researcher briefly described the meaning of each role to all
students at the beginning of class. Students were encouraged to help their group members and
discuss their ideas on how to solve a problem. Apart from the difference in one of the roles used,
the set up and structure of groups was the same in the control and treatment groups.
88

The instructional sequence followed in the control group was different to that used in the
treatment group. In the control group, the study of a probability topic began with looking at the
connection between the probability topics previously learned and the probability topic to be
studied that week. This was followed by a study of probability terminology and theoretical
probability on the topic of that week using a lecture format which included looking at the
bidirectional relationship between experimental and theoretical probability as a whole class using
examples from the textbook. This was different than the instructional sequence followed in the
treatment group in which students first worked in small groups on a probability activity prior to
formal instruction on a topic.
Following the above instructional sequence, the instructor solved a few probability
problems from the textbook on the particular topic studied. Then, students were placed in groups
and were asked to solve problems from the textbook. The instructor circulated around the
classroom during this time. Students submitted a copy of their group’s solutions to the instructor
at the end of class and feedback was provided to them during the next class session. Finally,
students were assigned 2-3 problems from the textbook to work on individually as homework.
These were corrected by the instructor and handed back to the students.
4.4.3. Commonalities between Treatment and Control Groups
In all course sections (control and treatment groups), students were provided with
opportunities to work with their classmates during class. Since, “students do not always enact the
roles as the teacher would like” (Anderson et al., 1997; Ross et al., 1996; as cited in Esmonde,
2009), the researcher/instructor circulated in the classroom as the students worked in their groups
and tried to monitor and encourage students to participate. In this manner, the instructor was also
able to respond to student questions as they worked in their groups. The students in the treatment
89

group were given opportunities to work cooperatively in small groups at the beginning of the
study of a probability topic through working on a probability activity, whereas the students in the
control group had such opportunities at the end of the study of a probability topic through
working on probability problems from the course textbook. In all sections, part of class time was
devoted to lecturing on probability terminology and theoretical probability as well as to wholeclass discussion.
In the two treatment sections as well as in the control group, group structuring was
determined by the instructor with students being placed in groups of 2-4 students. Member roles
were rotated between each group session so that students would have the opportunity to carry out
all roles by the end of the course. An attempt was made to include in each group students of
varying mathematical ability in order to create opportunities for scaffolding (Vygotsky, 1978; as
cited in Giraud, 1997). In placing students into groups, the students’ high-school mathematics
grade (upon entrance to the college), their pre-test score, and their English language skills were
taken into consideration.
In order to be able to gather information regarding students’ responses to the group
activities or problem sets as well as their reasoning and understanding of the probability
concepts, the conversations of students in both the treatment group and the control group were
audio-recorded. Since the researcher was also the course instructor, it would have been difficult
to be aware of the development of students’ reasoning and understanding of probability concepts
without a means of recording the students’ conversation. Overall, 19 audio recordings (10 from
the treatment group and 9 from the control group) were collected during the Spring 2010
semester as students worked either on an activity or problem set. Information relative to the
audio-recordings is provided in Table 4.2.
90

Table 4.2
Number Of Audio Recordings Collected In Each Group

Activity 1
Activity 2
Activity 3
Activity 4
Total

Treatment Group
# of Audio Tapes
2
2
3
3
10

Problem-set 1
Problem-set 2
Problem-set 3
Problem-set 4

Control Group
# of Audio Tapes
2
2
3
2
9

An audio recorder was provided to each group of students that was to have their conversations
recorded. At the end of the day, the audio recordings were downloaded to the researcher’s
personal computer which could be accessed through a password known only to the researcher.
Audio recordings were deleted from the recorders once they had been downloaded. The audio
recordings were transcribed during December 2011 and January 2012.
Although the official language of instruction at the college is English, since many
students have difficulties with the English language, instruction in all sections of MAT 201 in
spring 2010 was carried out in both English and Modern Greek. Through conversations with
colleagues and with the course coordinator at the college, it became known to the
researcher/instructor that students taking MAT 201 each year tended to have similar difficulties
with the English language. The instructor of MAT 201 in spring 2010 was a native Greek
speaker who is fluent in spoken and written English. Statistics and probability terminology were
taught to students in English but word problems were read to students in English and then
translated into Modern Greek. Furthermore, during lecture, whatever explanations, examples,
and procedures were provided in English (in written and verbal form) they were also given orally
in Modern Greek. The aim of this was to speak in the language that the students were more
91

comfortable with in an attempt to help them better understand the statistics and probability
concepts presented in class.
th

The final exam for MAT 201 took place on June 4 , 2010, it was graded by the
researcher/instructor and final course grades were submitted a week later. The envelope with the
signed consent forms was handed to the researcher during the second week of June 2010.
4.5. Instruments
Three instruments were developed for the study: a student background questionnaire, a
probability pre-test, and a probability post-test. All three were used in each of the three course
sections.
4.5.1. Student Background Questionnaire
The background questionnaire was administered during the first week of classes in all
sections of MAT 201. It consisted of 11 items and was used to specify such information as the
students’ age, Modern Greek and English competency, class level, major field of study, and prior
exposure to probability (see Appendix B). The students did not receive any points to count
towards their course grade for completing the background questionnaire.
4.5.2. Probability Pre-test
The probability pre-test was administered during the first week of classes (February 9

th

–

th

10 ) of the spring 2010 semester in all sections of MAT 201. It was used in order to identify
whether the students in the control group and the students in the treatment group had equivalent
initial probability knowledge. In turn, this helped in identifying the type of analysis that needed
to be carried out in order to address the research questions.

92

The pre-test comprised of a set of multiple-choice items from the web ARTIST
(Assessment Resource Tools for Improving Statistical Thinking) project and in particular from
the Probability Scale created by the ARTIST investigators (delMas et al., 2006; Garfield et al.,
2006), as well as multiple-choice items on probability from the TIMSS studies for grade 8
(TIMSS & PIRLS International Study Center, 1995, 2001, 2007, 2009). There exist 11 online
ARTIST topic scales each consisting of 7-15 multiple-choice items. The 11 topic scales are: data
collection, data representation, measures of center, measures of spread, normal distribution,
probability, bivariate quantitative data, bivariate catergorical data, sampling distributions,
confidence intervals, and significance tests. The ARTIST Probability Scale covers material that
is identified in the intended probability curriculum in Cyprus. Therefore, it was expected that the
students would be able to respond to these items. Moreover, since Cyprus participated in all
TIMSS activities for grade 8 through the years, some of the TIMSS items on probability for
grade 8 were included on the pre-test.
A total of 14 multiple-choice items made up the pre-test, eight from the web ARTIST
project (Items 1-6, 8 and 9) and six from the TIMSS studies (Items 7 and 10-14) (See Appendix
B). Table 4.3 provides a listing of the sources of the pre-test items along with information made
available for some of the TIMSS items regarding the international percentage of students
responding correctly to an item.
Some of these items underwent modifications based on feedback from experts and based
on participant responses in the pilot study. This feedback was also used to help reduce the
number of items on the initial version of the pre-test so that all items on the pre-test could then
be embedded on the final exam in spring 2010 as part of the post-test. Students were given 4%
towards their course grade for completing the pre-test.
93

Table 4.3
Pre-Test Items - Sources
Item(s)
1-6, 8, 9
7
10
11
12
13
14

Source
ARTIST Probability Scale
TIMSS 1995
(TIMSS & PIRLS International Study Center, 1995)
TIMSS 1995
(TIMSS & PIRLS International Study Center, 1995)
TIMSS 1999
(TIMSS & PIRLS International Study Center, 2001)
TIMSS 2007
(TIMSS & PIRLS, 2009)
TIMSS 2003
(TIMSS & PIRLS International Study Center, 2007)
TIMSS 2007
(TIMSS & PIRLS, 2009)

4.5.3. Probability Post-test
The post-test was embedded in the course final exam which was administered on June
th

4 , 2010 and was used to assess students’ knowledge and understanding of probability after
instruction. When constructing the probability post-test, the intended curriculum for MAT 201 –
Statistics I was considered. All items on the pre-test were embedded on the post-test. In addition,
the post-test included a multiple-choice item on discrete probability distributions and two freeresponse items: one on conditional probability and one on the Binomial distribution. The posttest is included in Appendix B.
Grading policy at the research site specified that the final exam for a course should count
for 70% of the course grade and should have duration of 3-4 hours. The final exam for MAT 201
in spring 2010 had duration of 3.5 hours. The post-test was the part of the final exam that

94

comprised of 15 multiple-choice items and 2 free-response items which examined the topics
under consideration in this study.
Several items, on the pre-test and on the post-test, required students to use theoretical
probability. For example, in items 4 and 5 students were provided with the sample space of
rolling two fair dice and were asked to determine equally likely and non-equally likely events.
Item 17 required students to determine the probabilities of events following the binomial
distribution and to compute the mean and variance of the distribution. Moreover, item 18
involved finding conditional probabilities.
Some items on both the pre-test and post-test, described experimental situations. For
example, item 1 dealt with buying lottery tickets and recognizing equally likely, independent
events. Item 3 involved flipping a fair coin and indicating the chance of getting a head on the
next flip after getting five consecutive heads. In addition, item 6 described a situation in which
the experimental approach would need to be used to estimate the probabilities of events.
Moreover, items 8 and 9 described experimental situations involving two containers filled with
various quantities of red and blue marbles, in which students needed to recognize equally likely
and non-equally likely events. Item 12 involved a probability situation of selecting marbles from
a bag without replacement and indicating what is the likely color of the next marble.
4.6 Data Analysis Plan
4.6.1 Quantitative Analysis
In order to be able to carry out quantitative analyses, student responses on the pre-test and
post-test items were coded. The pre-test comprised of multiple-choice items only and the codes
used in SPSS were: 0 for incorrect; 1 for correct; and 999 for missing data. The same codes were
used for the multiple-choice items on the post-test. For each free-response item on the post-test a
95

scoring rubric was created (See Appendix D). The codes used in this case were: 0 for incorrect; 1
for partially correct; 2 for completely correct; and 999 for missing data. The researcher consulted
with the Center for Statistics on the MSU campus and the members of the committee when
deciding which method was best to use in quantitatively analyzing the data.
Relative to the research questions, first a comparison of gain scores within each group
(control and treatment) from pre-test to post-test was performed in order to determine the effect
of each instructional method on students’ achievement in probability. That is, the pre-test scores
of the treatment group were compared to the post-test scores of the same group in order to
determine if the instructional treatment had an effect on students’ achievement. Also, the control
group’s pre-test and post-test scores were compared to determine if the instructional method used
in this group had an effect on students’ achievement. Second, a comparison of normalized gain
scores was performed in order to establish whether the instructional method used in the treatment
group (Instructional Method B) had a better effect on students’ achievement on probability than
the instructional method used in the control group (Instructional Method A). Moreover, a
comparison of post-test scores on the open-ended items and of post-test total scores was carried
out.
In each of these comparisons, descriptive measures were first computed for student
scores in the control and treatment groups and next, tests were performed to determine whether
the data followed the normal distribution. Normality tests helped determine the type of test
(parametric versus non-parametric) to be used when comparing student scores in the treatment
and control groups. In the cases that data followed the normal distribution, Welch’s test was used
to compare student scores in the treatment and control groups; this test does not assume equal
variances between two samples and so, it is a more general method to use than ANOVA. If data
96

did not follow the normal distribution, the Mann-Whitney non-parametric test was used to carry
out such comparisons. “The Mann-Whitney test is used for testing differences between means
when there are two conditions and different subjects have been used in each condition” (Field,
2000, p. 49). In the case of analyzing gain scores, if data did not follow the normal distribution,
then the Wilcoxon Signed-Ranks non-parametric test was used. “The Wilcoxon test is used in
situations in which there are two sets of scores to compare, but these scores come from the same
subjects” (Field, 2000, p. 54). Furthermore, items in which students performed substantially
worse on the post-test than on the pre-test were identified as this seemed to be an odd occurrence
after formal instruction on probability. Data on these individual items are presented in Chapter 5;
possible explanations for this decrease in performance are provided in Chapter 7.
In addition to an analysis of scores relative to achievement, a distractor analysis of
multiple-choice items was carried out in order to determine the effects of the instructional
treatment on students’ understanding of probability. As a first step, the percent-correct responses
on each multiple-choice item on the pre-test and post-test were computed in order to determine
how difficult students found these items to be. The correct response to each multiple-choice item
on the pre-test and post-test is provided in Table 4.4. The multiple-choice items that students
found to be less difficult on the post-test compared to the pre-test, as well as those that they
found to more difficult on the post-test compared to the pre-test were identified.
As a second step in the distractor analysis, the heuristic or misconception associated with
each distractor on each multiple-choice item was identified by the researcher. This was based on
information on heuristics and misconceptions that have been defined in the literature such as the
equiprobability bias, positive and negative recency, outcome approach and representativeness
heuristic. The list of item distractors and associated heuristics or misconceptions was then
97

examined by the researcher’s dissertation committee members and a meeting was held to discuss
any disagreements. Following this, a revised list of item distractors and associated heuristics or
misconceptions was created. The percentage of students that selected each distractor on the pretest as well as on the post-test was determined (see Appendix D). Then, item parts assessing the
same heuristic or misconception were grouped together and the mean percentage of students who
applied the particular heuristic or misconception on the pre-test and on the post-test in each
group was computed. Moreover, the heuristics or misconceptions for which the mean percentage
of students who applied them in each group increased or decreased were indicated.
Table 4.4
Content – Heuristic – Misconception Assessed By Each Multiple-Choice Item
Item
1

Correct Response
Content Assessed by Item
c - Student recognizes equally
likely independent events

2

c - Student expresses an
understanding of the
meaning of probability

3

b - Student recognizes equally
likely independent
outcomes
d - Student recognizes equally
likely events

4

5

b - Student correctly computes
probability using
combinatorial reasoning

6

b - Student expresses an
understanding of the Law
of Large Numbers

Incorrect Response
Heuristic/Misconception Assessed
a - Negative recency/Gambler’s
Fallacy
b - Positive recency
a - Misconception that ‘chance’
means ‘being lucky’
b - Deterministic approach to
probability
a - Negative recency/Gambler’s
Fallacy
c - Positive recency
a – Representativeness
b - Representativeness
c – Representativeness
a – Equiprobability bias
c - Inability to use sample space
provided to compute probability
d - Inability to use combinatorial
reasoning or sample space
provided
a - Equiprobability bias
c - Equiprobability bias

98

Table 4.4 (cont’d)

7

d - Student is able to use the
sample space to compute
the probability of the union
of events

8

b - Student understands that
probability of a compound
event is lower than or
equal to the probability of
its parts (i.e. the simple
events that make it up)
c - Student is able to i) use a
two-way frequency table to
compute probabilities and
ii) use the relative size of
outcomes to compare
probabilities
c - Student is able to use a
two-way frequency table to
compute the probability of
a compound event

9

10

a - Belief that the union (i.e. ‘or’) of
events means considering only
one of the events
b – P(A U B) = [P(A) + P(B)]/2
c - Belief that the union (i.e. ‘or’) of
events means considering only
one of the events
e - P(AUB) = P(A) + P(B)
a - Conjunction fallacy
c - Equiprobability bias

a - Reliance on absolute size instead
of relative size when comparing
probabilities
b - Reliance on absolute size instead
of relative size when comparing
probabilities
a - When computing the probability
of a compound event using a twoway table student divides by
column total
b - When computing the probability
of a compound event using a twoway table student divides by row
total
d - Misconception that since one item
is selected at random, the
numerator must be 1 and
probability = 1/number of
favorable outcomes

99

Table 4.4 (cont’d)
11

c - Student is able to use the
sample space to compute
the probability of an event

12

a - Student reasons about a
probabilistic situation that
involves selections
‘without-replacement’

13

b - Student recognizes nonequally likely outcomes
given in absolute/frequency
terms
a - Student expresses an
understanding of the
concept of probability
when information is given
in absolute/frequency terms

14

15
(Post-Test
only)

b - Student is able to find the
expected value of a
discrete probability
distribution

a - Misconception that since one item
is selected at random, the
numerator must be 1 and
probability = 1/number of
favorable outcomes
b - Misconception that since one item
is selected at random the
numerator is 1 and the
denominator is the characteristic
associated with the outcome
i.e. P(multiple of 3) = 1/3
d - Misconception that probability =
# of favorable outcomes/# of
remaining outcomes (i.e. use
odds)
b - Positive recency
c - Equiprobability bias
d - Ignores information about a
‘without-replacement’ situation.
Outcome approach.
a – Student is off task
c - Confuse ‘least likely’ with ‘most
likely’
d – Student is off task
b - Misconception that a larger
number of outcomes in sample
space means higher
probability
c - Equiprobability bias
d - Outcome approach
a - Ignores probability associated
with each outcome and thinks that
expected value is median value
c – Unweighted average of outcomes
d – Expected value / number of
categories

4.6.1.1 Initial Equivalence of Groups
In order to be able to investigate whether the two instructional methods under
consideration had a significant effect on students’ achievement and understanding of probability,
100

it was necessary to first identify whether the students in the three course sections had comparable
initial probability knowledge through an analysis of pre-test scores. The pre-test comprised of 14
multiple-choice items (See Appendix B). In SPSS, a correct response to an item on the pre-test
received a value of 1 and an incorrect response a value of 0. Missing data was coded as 999.
Therefore, the minimum score on the pre-test was 0 and the maximum 14.
Descriptive measures of pre-test scores by course section are provided in Table 4.5. The
morning treatment section had the highest mean and median whereas the evening treatment
section had the lowest mean and median. Also, the control group had the smallest degree of
variation from the mean whereas the evening treatment section had the highest variation from the
mean as well as the highest standard error.
Table 4.5
Descriptive Measures Of Pre-Test Scores By Course Section
N
Control
(Section B / morning section)
Treatment 1
(Section A / morning section)
Treatment 2
(Section C / evening section)

Mean

Median
9

Standard
Deviation
1.9

Standard Error
of the Mean
0.5

17

8.7

20

9.6

10

2.4

0.5

7

7.3

8

3.4

1.3

A further breakdown of the pre-test data by score gave rise to the following frequencies
as presented in Figure 4.1. This breakdown indicates that the mode for the control group was 9,
for the morning treatment section the modes were 9, 10 and 11 whereas for the evening treatment
section it was 6. When data from the three course sections were combined, the mode was 9.
Overall, 30 of the 44 participants received scores between 8 and 11 on the pre-test. Of these 30

101

participants, 11 were from the control group, 15 from the morning treatment section and 4 from
the evening treatment section.
Figure 4.1
Pre-Test Score Frequencies By Course Section
For Interpretation of the references to color in this and all other figures, the reader is referred to
the electronic version of this dissertation.

Following the computation of descriptive measures, an analysis was carried out to
identify whether the pre-test results followed a normal distribution. When the data from all three
sections was considered together, both the Kolmogorov-Smirnov and the Shapiro-Wilk tests for
normality resulted in p-values lower than 0.05 (0.005 and 0.023 respectively for each test). This
meant that the distribution of the sample was significantly different from a normal distribution.
102

Since the treatment occurred in two different sections (Sections A and C) which had
considerably different class sizes (20 versus 7 students respectively) and occurred at different
times during the day (morning versus evening), it was necessary to check for class effects prior
to carrying out analyses to determine whether the three course sections were initially equivalent
with respect to their knowledge of probability. So, an analysis was first carried out for the three
sections separately using a non-parametric method due to the small sample sizes of 20, 17 and 7
participants in Sections A, B and C respectively, and due to the fact that the sample data did not
follow a normal distribution. In particular, the Kruskal-Wallis test for three independent samples
was carried out using SPSS with codes assigned as follows: 0 for participants in the control
group i.e. Section B; 1 for participants in the first treatment section i.e. Section A (morning); and
2 for participants in the second treatment section i.e. Section C (evening). Following this, the
Kruskal-Wallis test was carried out for k = 2 independent samples. In the case of two
independent samples, two sections were compared at a time: control versus treatment1 (i.e.
Section B versus Section A); control versus treatment2 (i.e. Section B versus Section C); and
treatment1 versus treatment2 (i.e. Section A versus Section C). Table 4.6 shows the results of
these tests.

103

Table 4.6
Results Of The Kruskal-Wallis Test For k-independent Samples

k=3

Control
(Section B-morning)
Treatment1
(Section A-morning)
Treatment2
(Section C-evening)
p-value

Sample
Size
17

20.4

20

26.5

7

Mean Rank
k=2
k=2
Control
Control
vs.
vs.
Treatment1
Treatment2
16.1
13.3

16.3

21.5

k=2
Treatment 1
vs.
Treatment2

15.5
10.6

0.13

0.13

9.6

0.40

0.09

Since in all cases the p-value was higher than 0.05, there were no significant class effects. That
is, Section B (control) did not differ significantly from either Section A (treatment) or Section C
(treatment) with respect to pre-test scores. In addition, the two treatment sections (Section A and
Section C) did not differ significantly from each other.
Based on the above results, the data from the two treatment sections were then grouped
together to form one treatment group which was compared to the control group. The comparison
of the treatment versus control pre-test scores was performed using the Kruskal-Wallis nonparametric test for two samples since, as previously mentioned, the data differed significantly
from the normal distribution and sample sizes were small. Note that the Kruskal-Wallis test is
used in the place of a one-way ANOVA when the data is not normally distributed (Montgomery,
1997). In particular, the test resulted in a non-significant p-value of 0.37. In conclusion, the
participants in the control group had comparable initial probability knowledge to the students in
the treatment group.

104

4.6.2 Qualitative Analysis
For the purposes of analyzing students’ understanding of probability, six students were
selected (three from the treatment group and three from the control group) and their
conversations were fully transcribed: one student of high mathematical ability, one student of
moderate mathematical ability and one student of low mathematical ability from each group. The
selection criterion was students’ mathematics grade in the last year of high school. This
information was obtained from the background questionnaire (question 8; see Appendix B). In
order to perform this selection, students in the treatment group and in the control group were
placed into three categories: those who received a mathematics grade lower than or equal to 10;
those who received a mathematics grade between 11 and 15; and those who received a
mathematics grade between 16 and 20 in their last year of high school. Note that the maximum
grade one could receive in a high school subject is 20 with 10 being the passing grade. Next,
students in the treatment group and students in the control group were assigned a number and
one student from each category was randomly selected using the random generator function in
Excel. Data collected in the form of audio tapes were qualitatively analyzed.
The audio-taped student conversations were transcribed in December, 2011 and January,
2012 and then underwent a qualitative analysis during February, 2012. The researcher, along
with the help of a hired transcriber, completed the transcriptions of the audio-taped data. The
hired transcriber was a Cypriot college senior student majoring in elementary education who had
completed two courses in mathematics education and a course on research methods, and who
was fluent in both Modern Greek and English. Recall that overall 19 audio recordings were
collected; 10 in the treatment group and 9 in the control group. The researcher transcribed the 9
audio recordings collected in the control group and the hired transcriber the 10 audio recordings
105

collected in the treatment group. Once transcriptions were completed, the researcher and the
hired transcriber exchanged audio tapes and their corresponding transcripts in order to check
each other’s work.
As mentioned in Chapter 2 the framework developed by Jones, Thornton, Langrall and
Tarr (1999) was used to analyze student conversations. This framework included six constructs
(sample space, experimental probability of an event, theoretical probability of an event,
probability comparisons, conditional probability, and independence) across four levels of
reasoning (Level 1: Subjective; Level 2: Transitional; Level 3: Informal Quantitative; and Level
4: Numerical) as specified in Table 2.1 in Chapter 2.
Once transcriptions of audio tapes were completed, the first step in the qualitative
analysis was to read the transcripts to become acquainted with the data. After the first reading of
the transcripts, the researcher read the transcripts a second time and noted all of the vignettes
which related to a particular construct under examination in this dissertation and which involved
participation of one of the six students under study. During a third reading of the transcripts, the
researcher mapped each of these vignettes to the levels of reasoning on the framework. Then, the
researcher read the transcripts once more to confirm her impressions and to check whether there
were any instances that she disagreed with her first attempt of mapping a vignette to a level of
reasoning (intra-rater reliability). No such instances were noted.
At this stage, the (same) hired transcriber was called to help with the qualitative analysis.
A part of one of the transcripts for Activity 1 was used as an example to demonstrate to the
transcriber how to analyze the data. Then, clean hard copies of all transcripts were provided to
him. Once the hard copies of the transcripts were returned to the researcher, they were checked
for agreement on i) the number of vignettes related to the constructs under study and ii) the
106

levels of reasoning associated with each vignette. Then, inter-rater reliability was computed. The
researcher and the second coder (transcriber) met once more to discuss and resolve any instances
of disagreement in coding. The repeated readings of the transcripts by the researcher, the analysis
carried out by a second coder (transcriber) and examination of students’ written work provided
triangulation in the analysis of the data.
In qualitatively analyzing the data, two types of codes were used. These codes are
specified in Table 4.7 and are used throughout this chapter.
Table 4.7
Types of Codes Used In Qualitative Analysis
Type of Code
Person Involved in Transcript Excerpt
THA – Treatment/High Ability Student
TMA – Treatment/Moderate Ability
TLA – Treatment/Low Ability
CHA – Control/High Ability
CMA – Control/Moderate Ability
CLA – Control/Low Ability
OGM – Other Group Member
IR – Instructor/Researcher

Level of Reasoning
L1 – Level 1: Subjective Reasoning
L2 – Level 2: Transitional Reasoning
L3 – Level 3: Informal Quantitative Reasoning
L4 – Level 4: Numerical Reasoning

A major part of the qualitative analysis was to decide the level of reasoning represented
by excerpts of student conversations. Table 4.8 provides a demonstration of the analysis methods
used in matching excerpts to levels of reasoning with regards to the construct of theoretical
probability of an event. This table indicates the detailed description of each level of reasoning as
specified in the Jones et al. (1999) framework, and presents examples of responses selected from
students’ oral justifications to questions in the activities (treatment group) or problem sets
(control group) as representing each of the four levels of reasoning particular to this construct.

107

Table 4.8
Theoretical Probability: Levels of Reasoning and Examples of Representative Excerpts
Level 1
Subjective
-predicts most/least likely
event on the basis of
subjective judgments
-recognizes certain and
impossible events

Activity 3: Treatment
OGM: Why did you place
chips on the high values?
TMA: OGM did. She will
never win.
….
TMA: What mistake could
you make to eliminate
your chances of winning?
Put all your chips on 15.
TMA: Mmmm, and 13, 14
as well.

Level 2
Transitional

Level 3
Informal
Quantitative
-predicts
-predicts
most/least likely
most/least likely
event on the basis events on the basis
of quantitative
of quantitative
judgments but
judgments
may revert to
-uses numbers
subjective
informally to
judgments
compare
probabilities
Activity 3:
Activity 1:
Treatment
Treatment
TMA: 8. No
OGM: Based on
chips.
the theoretical
OGM: Oh, come
probabilities that
on!
you have
TMA: 5. No
computed in 17
chips.
above, which
…
sums give the best
TMA: Come on
chance of winning
OGM, make it
the game you have
happen!
just played in your
…
groups?
OGM: Who won THA: Number 7.
the game? …
Which is six
Why do you think times. And we
he won?
should write the
TMA: Because he pairs that give the
was smart!
number 7. 1-6, 6…
1, 5-2, 2-5
TMA: And he
THA: 3-4, 4-3…
didn’t place chips
on numbers over
12.

Level 4
Numerical
-predicts most/least
likely events for 1-and
simple 2-stage
experiments
-assigns a numerical
probability to an event
(either a real probability
or a form of odds)
Activity 1: Treatment
THA: To get three is 2
out of 36 so 1 out of
18?
OGM: Yes.
THA: To come up four
it is 1-3, 2-2, 3-1? So
we have 3, 3 out of 36,
so 1 out of 12?
OGM: Yes.

The two excerpts provided as examples of L1 and L2 reasoning involved the participation
of the moderate ability student in the treatment group, and originated from Activity 3 on discrete
108

probability distributions. Note that initially the TMA student was able to recognize impossible
events (i.e. getting a sum of 13, 14 or 15 when two dice are rolled) and so, the particular excerpt
was labeled as representing L1 reasoning. At a later stage in the activity and once the game was
over, students were asked to specify their group’s winner and indicate why that person won.
Prior to providing a valid reason as to why the particular group member won (i.e. he had not
placed any chips on numbers over 12 whereas the rest of the group members did), the TMA
student made a couple of statements that were subjective in nature. In particular, the TMA
student i) seemed to believe that the result of rolling two dice depends on who rolls the dice
(“Come on OGM, make it happen!); and ii) stated that the OGM won “because he was smart!”.
Since, the TMA student reverted to subjective reasoning, the second excerpt taken from Activity
3 was labeled as representing L2 reasoning.
The excerpts provided as examples of L3 and L4 reasoning originated from Activity 3
and Activity 1 respectively in which students dealt with the sum of two dice. In Activity 3
students played a game in which they placed chips on a number line that had the numbers 1 – 15
written on it. Students rolled two dice on which the numbers 1-6 were written, and computed the
sum of the faces; on each roll, a student that had placed a chip over the number represented by
the sum removed one of his/her chips. The winner was the student who had his/her chips
removed first. In the excerpt provided in Table 4.8 as representing L3 reasoning, the THA
student correctly specifies that the sum of 7 gave the best chance of winning (i.e. most likely
event) since it had the largest number of outcomes associated with it. The THA student also
correctly listed the outcomes associated with this sum. However, the THA student did not use
numerical probabilities to make valid comparisons between the various sums and so, this student
exhibited L3 reasoning.
109

In the passage provided under L4 Reasoning, the high ability student attempted to find
the probability of i) getting a sum of three and ii) getting a sum of four when two dice are rolled.
In each case the student correctly considered the outcomes that result in each of these sums and
assigned a correct numerical probability to the event under consideration, therefore the excerpt
was labeled as representing L4 Reasoning.
4.7 Summary
This dissertation made use of a mixed-methods design that utilized a probability pre-test
and post-test along with audio-taped student conversations to examine students’ achievement and
understanding of probability. Table 4.9 presents a summary of the data sources associated with
each measure considered in this study.
Table 4.9
Data Sources and Associated Measures

Data Source
Pre-test scores
(Multiple-choice items,
Open-ended items, Total Scores)
Post-test scores
(Multiple-choice items,
Open-ended items, Total Scores)
Post-test multiple-choice
item distractor analysis
Audio recordings

Measure
Achievement Understanding





The results of the quantitative analysis relative to students’ achievement and
understanding of probability are presented in Chapter 5 whereas the results of the qualitative
analysis of audio recordings with respect to students’ understanding of probability are provided
in Chapter 6.
110

CHAPTER 5
STUDENTS’ ACHIEVEMENT AND UNDERSTANDING OF PROBABILITY RESULTS
OF QUANTITATIVE ANALYSIS

This chapter presents the quantitative results that reflect the effects of the instructional
treatment on students’ achievement and understanding of probability. The chapter includes i) a
comparison of gain scores within each group (treatment and control) and of normalized gain
scores between groups relative to the multiple-choice items that were common to the pre-test and
post-test; ii) a comparison of post-test scores on the open-ended items and iii) a comparison of
post-test total scores. Moreover, the effects of the instructional treatment on understanding are
presented through a distractor analysis of multiple-choice items. For the purposes of analysis
SPSS-PASW Statistics 18 was used.
5.1 Effects of Instructional Treatment on Achievement
Recall that the pre-test included 14 multiple-choice items that were common to the posttest. The pre-test and post-test were administered to three sections of an introductory statistics
course. Two of these sections served as the treatment group and one as the control group.
Participants in the control group had comparable initial probability knowledge to participants in
the treatment group (see Chapter 4).
5.1.1 Comparison of (Normalized) Gain Scores
5.1.1.1 Comparison of Gain Scores Within Each Group
In the comparison of gain scores, only the 14 multiple-choice items that were common to
the pre-test and post-test were considered. The post-test included an additional multiple-choice
item on finding the mean of a discrete probability distribution and two open-ended items which
111

were not included in the analysis of gain scores. In SPSS, each correct response to a multiplechoice item received a value of 1 and each incorrect response a value of 0. Therefore, for the
purposes of analysis of gain scores, the minimum score that could be achieved was 0 and the
maximum 14.
Descriptive statistics for the treatment and control groups indicate that while the mean
and median multiple-choice score for the treatment group increased from pre-test to post-test, the
mean and median for the control group decreased.
Table 5.1
Changes in Descriptive Statistics from Pre-Test to Post-Test
Control
Group
8.7
0.45
9

8.96
0.53
9

6.47
0.55
6

Pre-Test
Mean
Standard Error
Median
Post-Test
Mean
Standard Error
Median

Treatment
Group

9.93
0.40
10

Based on the results of the analysis, the median pre-test score on the multiple-choice
items was 9 in both the control and treatment groups. However, the median post-test score on the
multiple-choice items in the control group decreased by 3 score points whereas in the treatment
group it increased by 1 score point in comparison to the pre-test. Moreover, whereas 25% of
students in the control group scored lower than 7 on the pre-test multiple-choice items, 75% of
students in this group scored lower than 7 on the corresponding post-test items. That is, 50%
more students in the control group scored below 7 on the post-test multiple-choice items in
comparison to the corresponding pre-test items. In the case of the treatment group, 50% of
112

students scored below 9 on the pre-test multiple-choice items whereas 25% scored below 9 on
the corresponding post-test items. Considering the above information relative to post-test
multiple-choice scores, 75% of students in the control group scored below 7 while 75% of
students in the treatment group scored above 9. Furthermore, 25% of students in the treatment
group scored above 12 on the post-test multiple-choice items whereas none of the students in the
control group achieved such scores.
Since the data of pre-test multiple-choice scores did not follow a normal distribution (see
Chapter 4), the Wilcoxon Signed-Ranks (non-parametric) test was used to examine the score
gains of students on the multiple-choice items. “The Wilcoxon test is used in situations in which
there are two sets of scores to compare, but these scores come from the same subjects” (Field,
2000, p. 54). In this case, the two sets of scores that were compared were the pre-test and posttest multiple-choice scores. The same students who took the pre-test also completed the post-test.
The control group data included 12 negative ranks meaning that 12 out of the 17 students in this
group received a lower score on the post-test multiple-choice items compared to the
corresponding pre-test items. Only two students (i.e. 12%) in the control group performed better
on the post-test multiple-choice items in comparison to the pre-test multiple-choice items. The
data from the treatment group included only 4 negative ranks whereas 14 out of the 27 students
(i.e. 52%) in this case received higher scores on the post-test multiple-choice items than on the
pre-test multiple-choice items.

113

Table 5.2
Wilcoxon Signed-Ranks Tests for Multiple-Choice Gain Scores

Negative Ranks
Positive Ranks
Ties
Total (i.e. N)
z-score
z-score basis
Significance

Control
Group
12
2
3
17
-2.52
Positive ranks
0.012

Treatment
Group
4
14
9
27
-1.71
Negative ranks
0.087

Relative to the z-scores generated by the analysis, these were significant only in the case
of the control group. The negative z-score (-2.52) in this case was based on the positive ranks,
meaning that student scores on the multiple-choice items moved in the opposite direction (i.e.
decreased) from pre-test to post-test. Since this z-score was significant (p = 0.012 < 0.05), this
means that the multiple-choice scores of students in the control group were significantly lower
on the post-test compared to the pre-test. In the case of the data from the treatment group, the
negative z-score (-1.71) was based on the negative ranks, meaning that student scores on the
multiple-choice items moved in the same direction (i.e. increased) from pre-test to post-test.
However, this z-score was not significant (p = 0.087 > 0.05), meaning that student scores on the
multiple-choice items of students in the treatment group did not differ (increase) significantly
from the pre-test to the post-test.
Overall, the results of the Wilcoxon Signed-Ranks tests indicated that i) students in the
control group performed significantly lower on the post-test multiple-choice items and ii)
students in the treatment group did not perform significantly different on the post-test multiplechoice items compared to the corresponding pre-test items.
114

In addition to the above results relative to score gains, figures were created using Excel
that demonstrate the percent-correct responses on each multiple-choice item on both the pre-test
and post-test (Figure 5.2 and Figure 5.3 for the control and treatment groups respectively).
Furthermore, Excel was used to construct a figure which shows the differences in percent-correct
responses between the pre-test and post-test for each multiple-choice item (Figure 5.4 on score
gains).
As can be seen in Figure 5.1, in the case of students in the control group, there was an
increase in the percent-correct responses on some items, a reduction in percent-correct responses
on other items while the percentage of correct responses remained stable in the case of two
items. In particular, in the control group gain scores were as follows:
Increase in percent-correct: Items 5, 7, and 10.
Percentage increase: 5.9%, 5.8% and 5.9% respectively.
Stable percent-correct: Items 2 and 9.
Decrease in percent-correct: Items 1, 3, 4, 6, 8, 11, 12, 13 and 14.
Percentage decrease: 17.6%, 5.9%, 52.9%, 11.7%, 11.7%, 58.8%, 5.9%, 41.1%, and
29.4% respectively.
These results indicate that for a majority of items, students in the control group had a
lower performance on the post-test compared to their performance on the pre-test. Moreover, all
of the percentage descreases are higher compared to the percentage increases. This is in
agreement with the Wilcoxon test results which indicated that the post-test multiple-choice
scores of students in the control group were significantly lower than their multiple-choice scores
on the pre-test. In particular, the most substantial descreases in percent-correct responses were
with regards to Items 4, 11 and 13. A discussion of the possible explanations for these
115

considerable percent-correct descreases is provided in Chapter 7. Note that in the case of Item 4,
responses exhibited a ceiling effect since all students in the control group provided the correct
answer to this item on the pre-test and so, it was not surprising that percent-correct responses
decreased on the post-test.
Figure 5.1
Control Group % Correct Responses on Pre/Post-Test Multiple-Choice Items

As Figure 5.2 shows, students in the treatment group demonstrated an increase in
percent-correct responses on some items, a reduction in percent-correct responses on other items
while the percent-correct responses remained stable in the case of three items.
Increase in percent-correct: Items 1, 3, 5, 7, 10 and 11
116

Percentage increase: 7.4%, 25.9%, 22.3%, 18.6%, 48.2% and 3.7% respectively.
Stable percent-correct: Items 8, 12 and 14.
Decrease in percent-correct: Items 2, 4, 6, 9 and 13.
Percentage decrease: 11.1%, 7.4%, 3.7%, 3.7% and 3.7% respectively.
Therefore, students in the treatment group had a percent-correct increase on six multiple-choice
items and a percent correct decrease on five multiple-choice items. Moreover, most of the
percentage increases were higher than the percentage decreases. The most substantial increases
in percent-correct responses were with regards to Items 3, 5, and 10. Note that, similar to the
results in the control group, the percent-correct responses on Item 8 remaimed low on both the
pre-test and post-test, indicating that this was an impossible item for students.

117

Figure 5.2
Treatment Group % Correct Responses Pre/Post-Test Multiple-Choice Items

Table 5.3 provides an additional demonstration of what has been previously discussed on
the basis of Figures 5.2 and 5.3. The results shown in Table 5.3 indicate that students in the
control group had a positive gain on three multiple-choice items; stable percent-correct responses
on two multiple-choice items and a negative gain on nine multiple-choice items. Specific to the
treatment group, students had a positive gain on six multiple-choice items; stable percent-correct
responses on three multiple-choice items and a negative gain on five multiple-choice items.

118

Table 5.3
Changes in Percent-Correct Responses on Multiple-Choice Items

Item
1
2
3
4
5
6
7
8
9
10
11
12
13
14

Control Group
% Change
-17.6
0
-5.9
-52.9
5.9
-11.7
5.8
-11.7
0
5.9
-58.8
-5.9
-41.1
-29.4

Treatment Group
% Change
7.4
-11.1
25.9
-7.4
12.3
-3.7
18.6
0
-3.7
48.2
3.7
0
-3.7
0

5.1.1.2 Comparison of Gain Scores Between Groups
In addition to the descriptive measures computed and the analysis of gain scores for each
of the two groups, a final piece of analysis was carried out with respect to gain scores which
directly addressed the research questions i.e. whether Instructional Method B (treatment) had a
better effect on students’ achievement on probability than Instructional Method A (control). In
order to carry out this part of the analysis, normalized gain scores were used (Bao, 2006; Hakes,
1998). That is, instead of simply subtracting pretest scores from posttest scores to produce the
raw gain, the raw gain was divided by the possible gain to get a ratio, g. This ratio was
computed for each individual student in each of the two groups (control and treatment) using the
following formula (Bao, 2006):

g

posttest score  pretest score
max imum score  pretest score
119

Based on this formula, the value of the normalized gain would, at most, be equal to 1. A
problem is caused when a student scores perfectly on the pre-test since in that case, the
denominator in the formula becomes zero (undefined). In this study, only one student (in the
treatment group) received a perfect score on the pre-test multiple-choice items so, a negligible
amount of data was lost due to this effect. In particular, the data collected from this student was
not used when carrying out tests to compare the normalized gain scores between the two groups.
Positive as well as negative normalized gains were possible. A student exhibited a
negative normalized gain score when the score received on the multiple-choice items was lower
on the post-test than on the pre-test. Particular to this study, 12 students in the control group and
4 students in the treatment group demonstrated negative normalized gain scores. The mean
normalized gain in the control group was -0.64 (standard error 0.23) whereas in the case of the
treatment group it was 0.066 (standard error 0.11).
Normality tests were then carried out to determine whether the data followed the normal
distribution. Both the Kolmogorov-Smirnov (p = 0.011 and p = 0.000 for the control and
treatment groups respectively) and the Shapiro-Wilk tests (p = 0.032 and p = 0.000 for the
control and treatment groups respectively) resulted in p-values lower than 0.05, indicating that
the data differed significantly from the normal distribution.
Subsequently, the Mann-Whitney non-parametric test was used to compare the
normalized gain scores of the control and treatment groups. The test resulted in a p-value of
0.001 (< 0.05) indicating that the normalized gain scores of the treatment group were
significantly different from the normalized gain scores of the control group. Note than in the
Mann-Whitney test, scores are ranked from lowest to highest. The test generated a mean rank of
14.41 for the control group and a mean rank of 26.96 for the treatment group. This means that
120

the control group had a bigger number of lower normalized gain scores as compared to the
treatment group. Likewise, the treatment group had a bigger number of higher normalized gain
scores in comparison to the control group. In summary, Instructional Method B (treatment) had a
significantly better effect on students’ achievement on probability than Instructional Method A
(control).
5.1.2 Comparison of Scores on Additional Post-Test Items
Apart from the 14 multiple-choice items involved in the analysis of gain scores, the posttest included an additional multiple-choice item on discrete probability distributions and two
open-ended items: i) Problem 16 on simple and joint probabilities, conditional probability and
independence and ii) Problem 17 on the binomial distribution (see Appendix B).
With regards to the multiple-choice item on discrete probability distributions (i.e.
problem 15 on the post-test; see Appendix B), 23.5% of students in the control group and 48.1%
of students in the treatment group responded correctly.
For the purposes of comparing the open-ended item scores of the control and treatment
groups, scoring rubrics were created (see Appendix C) which allotted numerical values to student
responses. Based on these rubrics, responses were coded by receiving a numerical value of 0
(incorrect), 1 (partially correct) or 2 (completely correct). This resulted in the first open-ended
item (Problem 16 on the post-test) receiving 18 points and the second open-ended item (Problem
17 on the post-test) receiving 6 points. Therefore, the minimum score on the open-ended items
that a student could receive was 0 and the maximum 24.
Descriptive statistics for the data generated by the two open-ended items are presented in
Table 5.4. In particular, the mean for the treatment group was 17.48, the standard error of the
mean was 0.77, and the standard deviation 3.98 (see Table 5.4). In comparison to the descriptive
121

statistics generated for the control group, the treatment group had a higher mean (12.24 for the
control group), a lower standard error (1.34 for the control group) and a lower standard deviation
(5.51 for the control group) (see Table 5.3). That is, the mean of the treatment group on the posttest open-ended items was 5.24 points higher than the mean of the control group. In addition, the
scores of the control group on the open-ended items had a higher variation from the mean in
comparison to the scores of the treatment group.
Table 5.4
Descriptive Statistics for the Open-Ended Item Scores
Statistic
N
Mean
Standard Error
Median
Standard Deviation

Control Group
17
12.24
1.34
14
5.51

Treatment Group
27
17.48
0.77
18
3.98

Based on the results, the control group had a median of 14, meaning that half of the
students in the control group received a score lower than 14 (out of 24) on the post-test openended items. However, with the exception of four students in the treatment group, the remaining
23 students in this group scored higher than 14. That is, 85% of the students in the treatment
group received a score higher than 14 on the post-test open-ended items. Moreover, whereas
75% of students in the control group scored lower than 16, only 25% of students in the treatment
group scored lower than 16 on the post-test open-ended items.
Following the computation of descriptive statistics, tests were carried out to determine
whether the data from the post-test open-ended items were normally distributed. Both the
Kolmogorov-Smirnov and the Shapiro-Wilk tests resulted in non-significant p-values for the
control group data (i.e. 0.11 and 0.3 respectively) and in significant p-values for the treatment
122

group data (i.e. 0.04 and 0.001 respectively). This meant that the distribution of the sample in the
case of the control group was not significantly different from a normal distribution but in the
case of the treatment group it was significantly different.
Subsequently, the Mann-Whitney non-parametric tests were used to compare the post-test
results from the open-ended items in the control and treatment groups. This resulted in a p-value
of 0.001 (< 0.05) indicating that the scores of the treatment group on the post-test open-ended
items were significantly different from the scores of the control group. Note that in the MannWhitney test, scores are ranked from lowest to highest. The test generated a mean rank of 14.15
for the control group and a mean rank of 27.76 for the treatment group. This means that the
control group had a bigger number of lower scores as compared to the treatment group.
Likewise, the treatment group had a bigger number of higher scores in comparison to the control
group.
In summary, students in the treatment group performed better on the additional post-test
multiple-choice item on discrete probability distributions. Moreover, based on the results of the
Mann-Whitney non-parametric test, the achievement of students in the treatment group on the
post-test open-ended probability items was significantly higher than the achievement of students
in the control group on the corresponding items. Therefore, Instructional Method B (using
lectures and small-group cooperative learning sessions that involve the use of activities that
generate real data) was successful in producing significantly higher achievement scores on the
post-test open-ended items compared to Instructional Method A (using lectures and small-group
cooperative learning sessions during which students solved probability problems).

123

5.1.3 Comparison of Post-Test Total Scores
As previously mentioned, a student could receive a maximum score of 15 on the
multiple-choice items and a maximum score of 24 on the open-ended items on the post-test.
Thus, overall, the maximum possible score on the post-test was 39.
Descriptive statistics for the data generated by the post-test items are presented in Table
5.5. Particular to the treatment group the mean was 27.89, the standard error of the mean was
1.01, and the standard deviation 5.26. In comparison to the descriptive statistics generated for the
control group, the treatment group had a higher mean (18.94 for the control group), a lower
standard error (1.61 for the control group) and a lower standard deviation (6.64 for the control
group). That is, the mean of the treatment group was 8.95 score points higher than the mean of
the control group. In addition, the post-test scores of the control group had a higher variation
from the mean in comparison to the post-test scores of the treatment group.
Table 5.5
Descriptive Statistics for the Open-Ended Item Scores
Statistic
N
Mean
Standard Error
Median
Standard Deviation

Control Group
17
18.94
1.61
20
6.64

Treatment Group
27
27.89
1.01
29
5.26

Based on the results, the median for the control group was 20 whereas the median for the
treatment group was 29. Moreover, in the control group 75% of the data was below 22 whereas
in the treatment group 74% of the data was above 27 (7 students in the treatment group received
scores 27 or lower).

124

Following the computation of descriptive statistics, tests were carried out to determine
whether the post-test data were normally distributed. Both the Kolmogorov-Smirnov and
Shapiro-Wilk tests resulted in p-values higher than 0.05 in the case of the control group (i.e. 0.2
and 0.9 respectively for each test) and in p-values lower than 0.05 in the case of the treatment
group (i.e. 0.00 for both tests). Based on these results, the data for the control group did not
differ significantly from the normal distribution whereas the data for the treatment group
deviated significantly from the normal distribution.
Subsequently, non-parametric tests were used to compare the total post-test scores of the
control and treatment groups. In particular, the Mann-Whitney test was used. This resulted in a
p-value of 0.00 (< 0.05) indicating that the post-test scores of the treatment group were
significantly different from the post-test scores of the control group. The test generated a mean
rank of 12.53 for the control group and a mean rank of 28.78 for the treatment group. This means
that the control group had a bigger number of lower scores compared to the treatment group.
Likewise, the treatment group had a bigger number of higher scores compared to the control
group.
In summary, the post-test achievement of students in the treatment group was
significantly higher than the post-test achievement of students in the control group. Therefore,
Instructional Method B was successful in producing significantly higher post-test scores
compared to Instructional Method A.
5.2 Effects of Instructional Treatment on Understanding
As mentioned in Chapters 1 and 2, the meaning of understanding given by Thompson and
Saldanha (2003) of “assimilation to a scheme” is used in this study. This allows for correct as
well as incorrect or inappropriate understandings (i.e. misconceptions) people may have. In this
125

study, students’ understanding of probability includes both appropriate as well as inappropriate
understandings. Therefore, two things were specified for each multiple-choice item on the pretest and post-test used in this study with regards to probabilistic understanding:
i) The goal of the item (correct choice). That is, the probability content or concept that the
item addressed and the student tried to understand and
ii) The heuristic or misconception associated with each distractor (incorrect choice) and
which the student might have applied on a particular item.
In this dissertation, students’ understanding of probability is measured through:
i) a distractor (quantitative) analysis of student responses to the multiple-choice items on
the pre-test and post-test (Chapter 5) and
ii) a qualitative analysis of audio-taped conversations as students worked in groups on
activities (treatment) or problem sets (control) (see Chapter 6).
5.2.1 Multiple-Choice Items: Content Assessed and Difficulty Level
As a first step in the distractor analysis, the percent-correct responses for each multiplechoice item were computed to determine how difficult the 44 students who completed the pretest and post-test found these items to be. If less than 30% of participants provided the correct
response to an item then participants found the item to be difficult. If 30-80% of participants
provided the correct response to an item then participants found the item to be of moderate
difficulty, and if more than 80% of participants responded correctly then participants found the
item to be easy. Table 5.6 provides the percent-correct responses for each multiple-choice item
on the pre-test and post-test used in this study (see also Appendix E), along with the difficulty
level exhibited by students.

126

As Table 5.6 shows, there was a change in the difficulty level exhibited by students in 11
of the 14 multiple-choice items that were common to the pre-test and post-test. In particular,
student difficulty level changed from Moderate to Easy in the case of 2 items, whereas in the
case of 6 items it changed from Easy to Moderate. In addition, student difficulty level changed
from Difficult to Moderate in the case of 2 items whereas for 1 item it changed from Moderate to
Difficult. Moreover, the difficulty level exhibited by students in Items 1, 8, and 9 remained
stable while students found item 15, which was only included on the post-test, to be difficult.
Table 5.6
Percent-Correct Responses on Multiple-Choice Items (N = 44)
Item
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Pre-Test
% Correct
86.4
84.1
68.2
93.2
27.3
31.8
29.5
15.9
65.9
50
88.6
79.5
81.8
81.8

Student Difficulty
Level
Easy
Easy
Moderate
Easy
Difficult
Moderate
Difficult
Difficult
Moderate
Moderate
Easy
Moderate/Easy
Easy
Easy

Post-Test
% Correct
84.1
77.3
81.8
68.2
43.2
25
43.2
11.4
63.6
81.8
68.2
77.3
63.6
70.5
38.6

Student Difficulty
Level
Easy
Moderate
Easy
Moderate
Moderate
Difficult
Moderate
Difficult
Moderate
Easy
Moderate
Moderate
Moderate
Moderate
Difficult

Table 5.7 indicates the content assessed by the four items (Items 3, 5, 7, and 10; see
Appendix B) in which there was a positive shift in the difficulty level exhibited by students i.e.
Moderate to Easy or Difficult to Moderate as well as the percentage change in correct responses
in the control and treatment groups. Notice that in the case of the control group, changes in
127

percent correct responses on these four items varied from -5.9% to +5.9% whereas in the
treatment group such changes varied from 18.6% to 48.2%. The highest overall increase in
percent correct responses occurred in the treatment group with regards to Item 10 which involved
using a two-way frequency table to compute the probability of a compound event. In the case of
the control group the highest increase in percent correct responses was at 5.9% in Item 5
(computing probability using combinatorial reasoning) and in Item 10. A breakdown of percentcorrect responses to these multiple-choice items on the pre-test and post-test for each course
section is provided in Appendix E.
Table 5.7
Multiple-Choice Items with Positive Shift in Student Difficulty Level
Item

Content Assessed

Change in
Difficulty Level
From Moderate
to Easy

3

Recognize equally likely
independent outcomes

5

Compute probability using
combinatorial reasoning

From Difficult
to Moderate

7

Use sample space to
compute probability
of union of events
Use a two-way frequency
table
to compute probability
of a compound event

From Difficult
to Moderate

10

From Moderate
to Easy

Change in
Percent Correct
Overall: + 13.6%
Control: -5.9%
Treatment:
Overall: + 15.9%
Control: +5.9%
Treatment: +22.3%
Overall: + 13.7%
Control: +5.8%
Treatment: +18.6%
Overall: + 31.8%
Control: +5.9%
Treatment: +48.2%

Table 5.8 specifies the content covered in the seven multiple-choice items (Items 2, 4, 6, 11, 12,
13, and 14) in which there was a negative shift in the difficulty level exhibited by students i.e.
Easy to Moderate or Moderate to Difficult as well as the percentage change in correct responses
in each group relative to these items. Notice that the highest overall decrease (-25%) in percent
128

correct responses occurred in Item 4 which involved recognizing equally likely events, while the
second highest overall decrease (-20.4%) occurred in Item 11 in which students were asked to
use the sample space to compute the probability of an event. These were the two items in which
the control group’s percent correct responses had the biggest decreases (-52.9% in Item 4 and 58.8% in Item 11). In the case of the treatment group the highest loss in percent correct
responses was in Item 2 which related to the meaning of probability. A breakdown of percentcorrect responses on every multiple-choice item on the pre-test and post-test in each course
section is provided in Appendix E.
Table 5.8
Multiple-Choice Items with Negative Shift in Student Difficulty Level
Item

Content Assessed

Change in
Difficulty Level
Easy to Moderate

2

Understanding of the
meaning of probability

4

Recognition of
equally likely events

6

Understanding of the
Law of Large Numbers

11

Use sample space to compute
the probability of an event

Easy to Moderate

12

Reason about a
‘without-replacement’
probabilistic situation
Recognize non-equally likely
outcomes given in
absolute/frequency form
Understand concept of
probability when information
is given in absolute/frequency
terms

Easy to Moderate

13

14

Easy to Moderate

Moderate to Difficult

Easy to Moderate

Easy to Moderate

129

Change in
Percent Correct
Overall: -6.8%
Control: 0%
Treatment: -11.1%
Overall: -25%
Control: -52.9%
Treatment: -7.4%
Overall: -6.8%
Control: -11.7%
Treatment: -3.7%
Overall: -20.4%
Control: -58.8%
Treatment: 3.7%
Overall: -2.2%
Control: -5.9%
Treatment: 0%
Overall: -18.2%
Control: -41.1%
Treatment: -3.7%
Overall: -11.3%
Control: -29.4%
Treatment: 0%

5.2.2 Multiple-Choice Items: Distractor Analysis
As previously mentioned, for each multiple-choice item on the pre-test and post-test, the
heuristic or misconception associated with each distractor was specified (see Table 4.4 in
Chapter 4). A table is included in Appendix E that indicates the percentage of students in each
course section that selected each distractor on each multiple-choice item on the pre-test as well
as on the post-test. For the purposes of distractor analysis, item parts assessing the same heuristic
or misconception were grouped together. The percentages of students selecting the particular
distractor on each of the grouped items were added up and the mean percentage of students
applying the particular heuristic or misconception on the pre-test and post-test in each group was
computed. Table 5.9 provides this information for heuristics or misconceptions that have been
defined in the literature such as the equiprobability bias, positive and negative recency, outcome
approach, representativeness, and use of absolute size when computing probabilities.
As the results in Table 5.9 show, in the case of the control group, the mean percentage of
students who applied the aforementioned heuristics or misconceptions either increased or
remained the same from the pre-test to the post-test. In particular, the mean percentage of
students who applied the outcome approach remained stable at 5.9%; the same was true with
regards to using the absolute size instead of relative size when computing probabilities
(percentage remained stable at a high 20.6%). In addition, there was a slight increase in the mean
percentage of students who applied the equiprobability bias which remained high on both the
pre-test and the post-test (28.4% and 29.4% respectively). Moreover, there was a small increase
in the mean percentage of students who applied the negative recency (by 2.9%) and the positive
recency (by 3.9%). The biggest change was in the application of the representativeness heuristic
(17.7% increase).
130

Unlike the results of the control group, the mean percentage of students in the treatment
group who applied the heuristics or misconceptions listed in Table 5.9 mostly decreased. In
particular, the mean percentage of students who used the equiprobability bias, negative recency,
positive recency and outcome approach decreased by 3.4%, 9.25%, 4.9% and 1.85%
respectively. Only in the case of the representativeness heuristic and use of absolute size instead
of relative size when computing probabilities the mean percentage increased by 3.7% and 2%
respectively.
Table 5.9
Mean Percentage Of Students Applying Particular Heuristics or Misconceptions
Heuristic or
Misconception
Equiprobability Bias
Negative Recency
Positive Recency
Outcome Approach
Representativeness
Use absolute size
instead of relative size
when computing
probabilities

Item
Distractor
5a, 6a, 6c,
8c, 12c, 14c
1a, 3a
1b, 3c, 12b
12d, 14d
4a, 4b, 4c
9a, 9b

Control

Treatment

Pre-Test
28.4

Post-Test
29.4

Pre-Test
20.4

Post-Test
17

11.8
5.9
5.9
0
20.6

14.7
9.8
5.9
17.7
20.6

11.1
8.6
3.7
2.5
11.1

1.85
3.7
1.85
6.2
13.1

Next, Table 5.10 considers students’ responses relative to other types of misconceptions
represented by item distractors. With regards to the first misconception listed in Table 5.10, the
mean percentage of students who believed that a larger number of outcomes in the sample space
implies a higher probability of occurrence increased in the case of the control group (from 0% to
5.9%) and decreased in the case of two treatment group (by 3.7%). Second, the mean percentage
of students who applied division by an incorrect total when computing probabilities using a two131

way table increased by 3% in the case of the control group and by 11.1% in the case of the
treatment group. Regarding the misconception that probability = 1/number of favorable
outcomes when one item is selected at random, the mean percent of students who applied this
misconception remained stable at 5.6% in the case of the treatment group but increased by 23.6%
in the case of the control group.
Notice that three of the misconceptions listed in Table 5.10 relate to the union of events.
The mean percentage of students who applied the first of these misconceptions (i.e. the term ‘or’
means considering only one of the events) was higher in the case of the control group on both the
pre-test and post-test while in both groups the percentage of students who applied this
misconception slightly increased from pre-test to post-test. Similarly, the mean percentage of
students who applied the misconception that P( A  B) 

P( A)  P( B)
was higher and remained
2

stable at 29.4% in the case of the control group on both the pre-test and post-test whereas it
decreased by 7.4% in the case of the treatment group (from 14.8% to 7.4%). Unlike the first two
misconceptions relating to the union of events, the mean percentage of students who applied the
misconception P( A  B)  P( A)  P( B) decreased in the case of the control group (by 5.9%)
and increased in the case of the treatment group (by 3.7%).

132

Table 5.10
Mean Percentage of Students Who Applied Particular Misconceptions
Misconception

Larger
number
of
outcomes in sample space
means higher probability
Division by incorrect
total when computing
probability using a twoway table
Union of events (i.e. ‘or’)
means considering only
one of the events
P( A)  P( B)
P( A  B) 
2
P( A  B)  P( A)  P( B)
Selecting one item at
random means
prob. = 1 / number of
favorable outcomes

Item
Distractor

Control

Treatment

Pre-Test
0

Post-Test
5.9

Pre-Test
3.7

Post-Test
0

10a
10b

5.9

8.9

0

11.1

7a
7c

11.8

14.7

3.7

5.6

7b

29.4

29.4

14.8

7.4

7e
10d
11a

23.5
0

17.6
23.6

18.5
5.6

22.2
5.6

14b

5.3 Summary
In summary, based on the analysis of gain scores, the achievement of students in the
treatment group on the post-test multiple-choice probability items was significantly higher than
the achievement of students in the control group on the corresponding items. Second, based on
the results of the Mann-Whitney non-parametric test, the achievement of students in the
treatment group on the post-test open-ended probability items was significantly higher than the
achievement of students in the control group on the corresponding items. Moreover, the overall
post-test achievement of students in the treatment group was significantly higher than the posttest achievement of students in the control group. Therefore, Instructional Method B (using
lectures and small-group cooperative learning sessions that involve the use of activities that
133

generate real data) resulted in significantly better effects on students’ achievement on probability
than Instructional Method A (using lectures and small-group cooperative learning sessions
during which students solved probability problems).
Specific to the multiple-choice items gain scores, the results of the Wilcoxon SignedRanks tests indicated that i) students in the control group performed significantly lower on the
post-test multiple-choice items compared to the corresponding pre-test items; and ii) students in
treatment group did not perform significantly different on the post-test multiple-choice items
compared to the corresponding pre-test items. Further analysis indicated that the normalized gain
scores of the treatment group were significantly higher than the normalized gain scores of the
control group meaning that Instructional Method B had a significantly better effect on students’
achievement in probability than Instructional Method A.
Results relative to the multiple-choice items indicated that for a majority of the multiplechoice items, students in the control group had a lower performance on the post-test compared to
their performance on the pre-test. Moreover, all of the percentage descreases were higher
compared to the percentage increases in performance. This was in agreement with the Wilcoxon
test results which indicated that the post-test multiple-choice scores of students in the control
group were significantly lower than their multiple-choice scores on the pre-test. In particular, the
most substantial descreases in percent-correct responses were with regards to Items 4, 11 and 13.
A discussion of the possible explanations for these considerable percent-correct descreases is
provided in Chapter 7. In the case of the treatment group, students had a percent-correct increase
on six multiple-choice items and a percent correct decrease on five multiple-choice items.
Moreover, specific to students in the treatment group, most of the percentage increases were
higher than the percentage decreases in performance.
134

In addition to the quantitative analysis carried out to measure students’ achievement on
probability, a distractor analysis was carried out to measure students’ understanding of
probability. Initially, student responses revealed the difficulty level of each multiple-choice item
on the pre-test and post-test. Then, the mean percentage of students who applied heuristics and
misconceptions defined in the literature was computed.
In the case of the control group: i) the mean percentage of students who applied the
outcome approach remained stable at 5.9%; ii) the mean percentage of students who used the
absolute size instead of relative size when computing probabilities remained stable at a high
20.6%; iii) there was a slight increase in the mean percentage of students who applied the
equiprobability bias which remained high at 28.4% and 29.4% on the pre-test and post-test
respectively; iv) there was a small increase in the mean percentage of students who applied the
negative recency (by 2.9%) and the positive recency (by 3.9%); and v) the biggest change was in
the application of the representativeness heuristic (17.7% increase from the pre-test to the posttest).
Unlike the results of the control group, in the case of the treatment group, the mean
percentage of students who applied the heuristics or misconceptions listed in Table 5.9 mostly
decreased. In particular, the mean percentage of students who used the equiprobability bias,
positive recency and outcome approach decreased slightly (by 3.4%, 4.9% and 1.85%
respectively). The most substantial decrease occurred relative to the use of the negative recency
i.e. 9.25%. Only in the case of the representativeness heuristic and use of absolute size instead of
relative size when computing probabilities the mean percentage increased by 3.7% and 2%
respectively.

135

A further presentation of results relative to the measurement of students’ understanding
of probability is provided in Chapter 6 which presents the results of the qualitative analysis of
audio-taped conversations as students worked in small groups while completing probability
activities or sets of probability problems.

136

CHAPTER 6
STUDENTS’ UNDERSTANDING OF PROBABILITY
RESULTS OF QUALITATIVE ANALYSIS
The purpose of this chapter is to discuss the effects of the instructional treatment on
students’ understanding of probability through a qualitative analysis of results. Recall that
Chapter 5 provided a presentation of the effects of the instructional treatment on students’
achievement and understanding of probability through a quantitative analysis that included a
comparison of pre-test and post-test scores as well as a distractor analysis.
This chapter begins by providing information relative to the activities and problem sets
used in the treatment and control groups respectively. Following this is a discussion of the results
of the qualitative analysis specific to students’ understanding of basic probability concepts such
as sample space, experimental probability of an event and theoretical probability of an event.
Next, a section on students’ understanding of conditional probability and independence is
included. Then, the chapter presents students’ understanding of discrete probability distributions
and last, their understanding of the binomial distribution.
For the purposes of qualitative analysis, transcripts of audio-taped student conversations
were coded using the levels of reasoning described in the Jones et al. (1999) framework that was
presented in Chapter 2. Six students were selected and monitored, three from the treatment group
and three from the control group, and their conversations were fully transcribed: one student of
high mathematical ability, one student of moderate mathematical ability and one student of low
mathematical ability from each group. The selection was carried out based on students’
mathematics grade in the last year of high school as described in Chapter 4.

137

Recall that two types of codes were used when analyzing the transcript data: codes
depending on the person speaking in the transcript excerpt (e.g. THA – Treatment/High Ability
Student; CMA – Control/Moderate Ability Student) and codes based on the level of reasoning as
presented in the framework (e.g. L1 – Level 1: Subjective Reasoning).
6.1 Information on Activities and Problem Sets
Table 6.1 provides information regarding the four activities completed by students in the
treatment group and the four problem sets completed by students in the control group. Detailed
descriptions of the activities and problem sets can be found in Chapter 4: Methods whereas the
activities and problem sets are provided in Appendix C.
Note that in the cases of Problem Set 2 and Problem Set 4, students in the control group
only worked on one problem in groups. These problem sets originally consisted of two problems
on which students were asked to work in small groups. However, they ended up completing only
one problem in each of these two problem sets since i) they took 40 minutes or more to complete
one problem alone; ii) all groups required complete translation into Greek of the context
provided in the problems and iii) all groups seemed to find the problems quite challenging. Once
a copy of students’ group work was collected, the second problem that had originally been
assigned as part of group work was solved on the board using a whole-class discussion approach.

138

Table 6.1
Information On Activities And Problem Sets
Activity 1 (Treatment)
Average Duration
54 minutes
Concepts
Sample space; Experimental
Covered
probability of simple and joint
events; Theoretical probability of
simple and joint events; Probability
comparisons; Law of Large
Numbers.
Activity 2 (Treatment)
Average Duration
57 minutes
Concepts
Experimental and theoretical
Covered
probability of simple events;
Sample space; Conditional
probability; Independence.
Activity 3 (Treatment)
Average Duration
26 minutes
Concepts
Impossible events; Sample space;
Covered
Probability distribution of a
discrete random variable.

Activity 4 (Treatment)
Average Duration
51 minutes
Concepts
Sample space; Experimental and
Covered
Theoretical probability of joint
events; Law of Large Numbers;
Binomial Distribution.

Problem Set 1 (Control)
44 minutes
Sample space; Simple events, joint
events, complementary events;
Probability of intersection of events;
Probability of union of events.
Problems 4.2, 4.8 and 4.9 (See
Appendix C)
Problem Set 2 (Control)
40 minutes
Conditional probability;
Independence.
Problem 4.23 (See Appendix C)
Problem Set 3 (Control)
38 minutes
Expected value of a discrete random
variable; Sample space; Probability
distribution of a discrete random
variable.
Problems 5.3a and 5.4 (See Appendix C)
Problem Set 4 (Control)
43 minutes
Expected value and standard deviation
of a discrete random variable that
follows the binomial distribution;
Binomial distribution.
Problem 5.13

6.2 Inter-Rater Reliability
As mentioned in Chapter 4, transcripts were coded independently by the researcher of
this study and by a second rater. A hard copy of each transcript was provided to the researcher
and to the second coder who i) read the transcripts once to become acquainted with the data; ii)
read the transcripts a second time and noted all of the excerpts which related to a particular
construct under examination and which involved participation of one of the six students under
study and iii) mapped each of these excerpts to the levels of reasoning on the framework. Once
139

the coding of transcripts was completed, they were checked by the researcher for agreement on
the number of excerpts related to the constructs under study and the levels of reasoning
associated with each excerpt. The researcher and second coder met once more to discuss and
resolve any instances of disagreement in coding. The percent-agreement for each concept is
presented in Table 6.2.
Note that Table 6.2 does not include the concepts of discrete probability distribution and
binomial distribution. This is due to the fact that the framework does not accommodate for these
concepts. So, students’ level of reasoning as exhibited in excerpts associated with discrete
probability distributions (in Activity 3/Problem Set 3) and the binomial distribution (in Activity
4 /Problem Set 4) could not be directly coded using the framework. However, in Activity
3/Problem Set 3 and in Activity 4/Problem Set 4 students needed to make connections to basic
probability concepts; reasoning on such concepts is discussed in the framework. Excerpts in
which students made connections to basic probability concepts are accounted for in Table 6.2
when computing the inter-rater reliability. Any excerpts relative to the concepts of discrete
probability distributions and the binomial distribution in Activity 3/Problem Set 3 and in Activity
4/Problem Set 4 respectively in which students did not make connections to basic probability
concepts are discussed in sections 6.4 and 6.5 of this chapter.
Table 6.2
Percent-Agreement Between Coders of Transcript Excerpts
Concept
Sample Space
Theoretical Probability
Experimental Probability
Conditional Probability
Independence

Number of Excerpts
Associated with Concept
14
54
26
11
8
140

Percent-Agreement in
Coding of Reasoning Level
13/14 = 93%
44 / 54 = 81.5%
22 / 26 = 84.6%
9 / 11 = 81.8%
100%

6.3 Effects of Instructional Treatment on Understanding of Basic Probability Concepts
In order to examine the effects of the instructional treatment on students’ understanding
of basic probability concepts, the first four constructs included in the Jones et al. (1999)
framework i.e. sample space, experimental probability of an event, theoretical probability of an
event, and probability comparisons were considered when analyzing transcripts of student
conversations. The audio-taped conversations of the students monitored in the treatment and
control groups as they worked in small groups on all activities and problem sets were
qualitatively analyzed with respect to basic probability concepts.
6.3.1 Sample Space
Overall, the four activities used in the treatment group provided more opportunities for
students to express their reasoning regarding the concept of sample space than the problems in
the course textbook that students in the control group worked on. Particularly, in problem sets 1,
2, and 4, transcripts of student conversations revealed no instances during which the three
students in the control group reasoned about sample space.
As shown in Table 6.3 the high ability student in the treatment group (THA) exhibited
Level 4: Numerical Reasoning with regards to sample space across all four activities. The high
ability student in the control group (CHA) expressed reasoning relative to sample space in
Problem Set 3 only in which case that reasoning was at L4 as well. That is, both students were
able to apply a generative strategy that enabled a complete listing of outcomes. For example,
when students were asked at the beginning of Activity 1 which sum they believed is most
probable when two dice were rolled the THA student responded as follows:
16

THA: So, we should write the pairs first. 1-1…

17

THA and OGM together: 1-2, 1-3, 1-4, 1-5, 1-6
141

18

THA: 2-1, 2-2, 2-3, 2-4, 2-5, 2-6, 3-1, 3-2, 3-3, 3-4, 3-5, 3-6, 4-1, 4-2, 4-3, 4-4,

19

4-5, 4-6, 5-1, 5-2, 5-3, 5-4, 5-5, 6-1, 6-2, 6-3, 6-4, 6-5, 6-6. 12 is out of the

20

question, same with 2 because they only show up once.

In addition, in Activity 2, when students were asked to determine the number of outcomes when
three dice are rolled the THA student reasoned in the following manner:
80

OGM: What does it mean by how many outcomes?

81

THA: It could be 1-1-1; 1-2-1; got it?

82

OGM: So, how many are there?

83

THA: I don’t know.

84

OGM: Infinitely many?

85

THA: It can’t be infinitely many because a die has numbers on it up to 6.

86

OGM: Wait, I have my class handout from last time when we were looking at
two dice.

87

THA: If we have two dice then the number of outcomes is 36.

88

OGM: Yes, it’s 36.

89

THA: So, here it should be 6 x 6 x 6; 216.

Moreover, the THA student was able to construct the sample space for three free-throw attempts
of a basketball player in Activity 4 (See Appendix C).
With regards to Activity 3 and Problem Set 3 in which students dealt with discrete
probability distributions, both the THA student and the CHA student were able to identify that i)
there are 36 possible outcomes when two dice are rolled; ii) alternating faces count as different
outcomes (e.g. 1-2 and 2-1 are two different outcomes); and iii) the possible sums when two dice
are rolled are 2 to 12. In particular, the CHA student reasoned as follows:
142

76

IR: So, how many outcomes give a sum of 2?

77

CHA: 1

78

IR: To get a sum of 3, how many outcomes give us that?

79

CHA: Three

80

IR: Which three?

81

CHA: 1-2, 2-1 and 3-0; oops there is no 0. So, only two. To get a sum of four

82

we have 2-2, 1-3, 3-1….

…..
89

CHA: For a sum of 5

90

OGM: 4-1, 1-4

91

CHA: 2-3

92

CHA: and 3-2

93

CHA: So, four.

94

OGM: For sum of six we have 1-5, 5-1

95

CHA: Yes

96

OGM: 3-3

97

CHA: Yes

98

CHA: 2-4, 4-2

99

CHA: Five. Oh I see how it goes. The next one should have six outcomes.

100

OGM: Then, 7, 8, 9, 10, 11

101

CHA: Wait, are you sure? Maybe we should write down the results to see if

101-102

that is the case. You know what we should do? We should add them up

102

to see if they are 36 altogether.
143

103

OGM: 1 plus 2, three; plus 3, six; plus 4, ten; plus 5, 15; plus 6, 21; … 66

104

CHA: Then it’s not like that. We are ok up to here. Lets see how many there

104-105

are for a sum of 7. We have 1-6, 6-1, 5-2…

As shown in Table 6.3, with regards to sample space, the moderate ability student in the
treatment group (TMA) exhibited L4 reasoning in Activity 1, L2 reasoning in Activity 2, and L4
reasoning once again in Activity 4. Particular to Activity 1, the TMA student was able to provide
a complete listing of the outcomes that generate each sum of two dice. In Activity 2, the TMA
student reasoned as follows in response to a question regarding the number of outcomes when
three dice are rolled:
106

IR: …When you have one die, how many possible outcomes are there?

107

TMA and OGM: 6

108

IR: When we have two dice?

109

TMA: 12

110

OGM: 36!

111

OGM: So, in this case it is 6 to the power of 3.

112

TMA: 36 times 6; 216.

The above excerpt provides evidence that the TMA student correctly indicated the number of
possible outcomes when one die is rolled, but only through the aid of another group member
could this student point out the number of possible outcomes when two or three dice are rolled.
In Activity 4 the TMA student “adopts and applies a generative strategy that enables a complete
listing of the outcomes” (Jones et al., 1999, p. 15) for three free-throw attempts of a basketball
player thus exhibiting L4 reasoning with regards to sample space:
11

TMA: How many outcomes are there? Well, he can make it, he can make it,
144

11-12

not make it. He is making three shots ok?

…
16

OGM: So, how many combinations there are?

17

TMA: How many possible outcomes.

18

OGM: He can make the basket three times; he can make it twice and lose once.

19

TMA: Yes, that’s right. To make it twice and lose once, that’s given, it’s BBN.

20

Lets start off by writing down, to make it three times.

21

OGM: It might be BBN ooops BNN.

22

TMA: BBB. BBN is given…

23

OGM: NNN

24

TMA: BNB…NBB…NNN

25

OGM: He could make it once and lose twice.

26

OGM: Or lose twice and make it once.

27

TMA: BNN…NNB

28

OGM: NBN

29

TMA: That’s right. …

The CMA student expressed the same sequence of reasoning (L4  L2  L4) relative to
sample space in Problem Set 3 only. The CMA student was initially able to adopt a generative
strategy to list the possible outcomes that relate to a particular sum when two dice are rolled but
later included outcomes that were not part of this sample space (i.e. 1-7 and 7-1 as outcomes
relating to a sum of 8 and 2-7 and 7-2 as outcomes relating to a sum of 9 when two dice are
rolled) thus exhibiting L2: Transitional Reasoning. However, the CMA student soon realized that
pairs which include numbers higher than 6 cannot be included as outcomes since a regular die
145

does not have a face value higher than 6 and continued to provide a complete listing of outcomes
associated with each sum. In this respect, the CMA student transitioned from L2 to L4 reasoning
with regards to sample space. The following excerpt demonstrates the aforementioned levels of
reasoning exhibited by the CMA student. Prior to this excerpt, students correctly indicated the
list of outcomes that resulted in sums of 2 to 7 when two dice are rolled, and were then involved
in the process of finding the number of outcomes that result in sums of 8 to 10.
128

OGM: It goes 1, 2, 3, 4, 5, 6 and then it will be 7, 8, 9, 10 … you’ll see. Eight.
1-7

129

CMA: 7-1, 4-4, 5-3, 3-5, 6-2, 2-6

130

OGM: There are more

131

CMA: That’s it! Seven

…
136

OGM: For nine … 1-8, 8-1

137

OGM: It will be eight of them, you’ll see.

138

CMA: 5-4, 4-5, 2-7, 7-2. Wait.

139

OGM: Let’s look at them again. 1-8, 8-1

140

CMA: Does a die have an 8?

141

OGM: Ooops I forgot.

142

CMA: We made a mistake before as well.

143

OGM: Yeah, you are right, we made a mistake.

144

CMA: Let’s start from 1 each time. 1-7, no. 2-6, ok.

145

OGM: 6-2

146

CMA: 3-5, 5-3, 4-4. That’s it. Five
146

…
149

OGM: Ok, for nine.

150

CMA: Nothing with 1.

151

OGM: 6-3, 3-6

152

CMA: 4-5, 5-4

153

OGM: Now it will go in reverse order.

154

OGM: Yes

155

CMA: Four

156

OGM: For ten. 5-5 only.

157

CMA: What? No, there are more.

158

OGM: 6-4, 4-6, 7-3, 3-7

159

CMA: There is no 7 on the die! 3 over 36

160

OGM: For eleven?

161

CMA: 5-6 and 6-5. 2 over 36. And the last one is 1 over 36.

As evident from this excerpt, for a while the CMA student, along with the other group members,
incorrectly considered face values that do not exist on a regular six-sided die but soon realized
their mistake. The CMA student then reverted to listing correct outcomes for sums of 8 to 10.
As shown in Table 6.3 the TLA student exhibited L4 reasoning in Activities 2 and 4 with
regards to sample space whereas the CLA student expressed L2 reasoning in Problem Set 3 only.
In Activity 2, the TLA student correctly indicated that there are 6 possible outcomes when one
die is rolled, 36 possible outcomes when two dice are rolled and “6 to the power of 3” (line 111)
possible outcomes when three dice are rolled. In Activity 4, the TLA student was able to use a

147

generative strategy to provide a complete listing of the eight possible outcomes for three freethrow attempts of a basketball player:
10

TLA: So, we need to find all the possible outcomes.

11

OGM: And this one counts as one of the outcomes right?

12

TLA: Yes.

13

OGM: So, BBN and then

14

TLA and OGM: BNB

15

OGM: NBB

16

TLA: Wait a second, you confused me. BBN, BNB, NBB … BNN

17

OGM: BBB

18

TLA: NNN

19

OGM: Is that it?

20

TLA: Is there anything else? NNB

21

TLA is repeating out loud the outcomes they already listed.

22

TLA: NBN. Did we include it? No.

As previously mentioned, the CLA student exhibited L2 reasoning in Problem Set 3 with regards
to sample space. This student was able to provide a partial set of outcomes associated with each
sum of two dice; the list was completed by the other group members.
Table 6.3 indicates the levels of reasoning and transitions between levels of reasoning
that the six students demonstrated relative to the concept of sample space. Overall, the activities
provided students in the treatment group with more direct as well as indirect opportunities to
reason about sample space compared to the problem sets used in the control group. In particular,
as indicated on Table 6.3, in problem sets 1, 2 and 4 students in the control group did not reason
148

at all relative to sample space. In addition, the reasoning exhibited by students in the treatment
group relative to sample space was more stable (i.e. it involved fewer transitions between levels
of reasoning) compared to reasoning exhibited by students in the control group. Moreover,
students mostly exhibited a high level of reasoning regarding the concept of sample space.
Table 6.3
Levels of Reasoning Exhibited by Students Relative to Sample Space

Activity
1
2
3

4

Treatment Group
THA TMA TLA
L4
L4
L4
L2
L4
L4

L4

L4

Problem Set
1
2
3

L4

Control Group
CHA CMA CLA

L4

L4
↓
L2
↓
L4

L2

4

6.3.2 Theoretical Probability of an Event
The framework by Jones et al. (1999) considers theoretical probability and probability
comparisons as two separate constructs. Since i) as stated in the framework, the construct of
probability comparisons involves making quantitative judgments (which are part of the construct
of theoretical probability) and ii) in many cases that students were asked to compute the
theoretical probability of an event in an activity or problem they were also asked to compare this
to the probability of another event, theoretical probability and probability comparisons were
considered together for the purposes of qualitative analysis in this study. Table 6.4 indicates the
levels of reasoning demonstrated by the six students relative to theoretical probability and
probability comparisons.

149

In Activity 1, the TMA and THA students that were monitored in the treatment group
transitioned from L2 reasoning to L4 reasoning while the TLA student remained stable at L2
reasoning with regards to theoretical probability. The first question posed in Activity 1 asked
students to state which sum they believed was most probable when two dice are rolled. Notice
that in Activity 1, the TLA student and the TMA student were in the same group.
2

TLA: 7

3

TMA and OGM: Number 7

4

TMA: And the reason is? It comes up through 6-1

5

TMA and TLA: 5-2 and 4-3

6

TMA: Number 7 … write it down dear.

The TLA and TMA students correctly responded that the sum of 7 is the most probable sum
when two dice are rolled. However, they did not consider all possible outcomes that result in a
sum of 7 (i.e. they did not consider alternating faces of dice). On the other hand, the THA student
correctly listed all possible outcomes when two dice are rolled (including alternating faces) but
incorrectly responded that the sum of 6 was the most probable sum when two dice are rolled
“because it is halfway through 2 and 12” (line 30). The reasoning of all three students at the
beginning of Activity 1 indicates that they made comparisons based on quantitative judgments.
However, their reasoning was not entirely correct, thus exhibiting L2 reasoning. The THA
student then continued in the attempt to convince the group members that the sum of 6 was the
most probable sum by pointing to the five outcomes that result in a sum of 6 and indicating that
there are only four outcomes that result in a sum of 5. Once again, the THA student exhibited L2
reasoning since this student did not consider the outcomes associated with each possible sum but
only focused on two of the possible sums. At a later stage, students were asked to compute the
150

theoretical probability of each sum from 2 to 12 when two dice are rolled using a pictorial
representation of the sample space that was provided to them in the activity (see Appendix C).
Both the TMA student and the THA student correctly computed these probabilities by
considering the outcomes associated with each sum and dividing by 36, thus exhibiting L4
reasoning with regards to theoretical probability (lines 226-247 and lines 272-303 in the
respective transcripts). Moreover, the TMA student added up the probabilities of the possible
sums to make sure that they resulted in 100% (lines 248-249).
Note that in Activity 1, the TMA student and the TLA student were in the same group.
Although the TLA student participated actively in the group work by i) providing very brief
responses, ii) converting fractions into percentages for the purposes of reporting probabilities and
iii) volunteering to construct one of the histograms, the oral responses provided by this student
were not elaborate enough to clearly indicate the level of reasoning carried with respect to
theoretical probability. A similar situation arose in the control group. The CHA student and the
CLA student were in the same group while working on Problem Set 1. Although the CLA
student participated actively in the group work by i) providing very brief responses and ii) asking
questions, the oral responses provided by this student were not elaborate enough to clearly
identify the reasoning level carried with respect to theoretical probability.
In the case of Problem Set 1, the three students that were monitored in the control group
initially exhibited L1 reasoning with respect to theoretical probability. The CLA student
remained stable at this level of reasoning, whereas the reasoning patterns followed by the CHA
and CMA students during group work were similar (CHA: L1  L4  L2; CMA: L1  L4 
L2  L4). The first two problems in Problem Set 1 required students to provide examples of
simple events, joint events and complementary events. All three students could identify
151

complementary events with ease. However, they had difficulties providing examples of simple
and joint events.
20

IR: You only need to provide one event but that event should be a simple one.

21

What does simple event mean?

22

CHA: That it has one characteristic.

23

IR: What is one characteristic that the ball could have?

24

CLA: It could be either red or green.

25

OGM: A simple event is the color.

26

CLA: One red or one green.

27

CHA: Event red or event green.

28

CLA: So, I will write Event A …

29

CHA: Not like that.

30

CLA: Red

31

CHA: or green. In part b it’s what OGM said before.

32

OGM: What did I say?

33

CHA: That if it’s not red then it will be green.

34

CLA: Here it says red so the answer is green ball.

35

CHA: Green balls … no ball. Only one will be chosen.

36

CLA: Oh come on! Details!

In their written work this group of students specified that “Event red or green balls” is a simple
event. On the other hand, the CMA student, who worked in a different group, could easily
express simple events verbally and in written form (i.e. “The ball is red”). When it came to
providing an example of a joint event the CHA student and the CMA student correctly identified
152

such an event both verbally and in written form. The CLA student had difficulty identifying such
an event but, with the help of the CHA student, understood how to use the contingency table
provided in the problem to identify two characteristics relative to the problem context which
when taken together may form a joint event (lines 88-128). The framework by Jones et al. (1999)
indicates that if a student “recognizes certain and impossible events” then he/she is exhibiting L1
reasoning with respect to theoretical probability. This could be extended to include other types of
events such as simple, joint, and complementary events. In that case, the three students in the
control group exhibited L1 reasoning at the beginning of Problem Set 1.
In the last problem in Problem Set 1 (i.e. problem 4.9; see Table 6.4), students were
asked to compute the probability of simple and joint events.
Table 6.4
Problem 4.9 (Levine, Krehbiel and Berenson, 2010)

U.S. Tax Code
Fair
Unfair
Total

Income Level
Less than $50,000 More than $50,000
225
180
280
320
505
500

Total
405
600
1,005

Referring to the contingency table, if a respondent is selected at random, what is the
probability that he or she
a. thinks the tax code is unfair?
b. thinks the tax code is unfair and makes more than $50,000?
c. thinks the tax code is unfair or makes more than $50,000?
d. Explain the difference in the results in (b) and (c).
The three students implemented L4 reasoning when working on finding the probability of a
simple event by assigning a correct numerical probability to it. When students were asked to
compute the probability of a joint event initially none of them could correctly carry out this
computation or identify the correct frequencies on the contingency table that needed to be used.
153

102
103

CMA: In part b we need the probability that ‘thinks the tax code is unfair and
makes more than 50000’. So, 320 over 500.

104

IR: Why over 500?

105

OGM: Over 1005?

106

IR: Whatever you think is correct.

107

OGM: 320 over 500.

This excerpt provides evidence that the CMA student used quantitative reasoning to compute the
probability of an event but this was carried out in an incorrect manner, thus exhibiting L2
reasoning. At the end of the group work, the same student realized that this computation must
have been incorrect. This realization came about when the probability computed in part b needed
to be used in part c. At that point, the CMA student indicated that “In part b it should be 320 over
1005 as well” (line 155). By assigning a correct probability to an event, the CMA student moved
to L4 reasoning with regards to theoretical probability. Similarly, the CHA student and the CLA
student were confused about which number should be used in the denominator of the
computation in part b.
196

CHA: 320

197

IR: Good.

198

CHA: Out of 600?

199

IR: Why out of 600?

200

CHA: Or over 1005? Or over 500? I am confused! Come on guys, help a bit!

201

CHA: 320 … 320 over 1005

202

CLA: Why 1005? Do we take the total?

203

CHA: That is the total but on the other hand it says ‘more than’ so we might need
154

204

to divide by 500.

205

OGM: It must be over 500.

206

CHA: CLA who is going to write this?

207

CLA: You can be the recorder.

208

CHA: So, I should be doing everything.

209

CLA: I don’t mind writing the answers but you should tell me what to write.

The written response that these students ended up providing was

600 280
320
which was


1005 1005 1005

correct. In their verbal communications students exhibited L2 reasoning; they were using
quantitative judgments to compute the probability of joint events; however, they were not doing
so in a completely correct manner. The CHA’s statement that the answer should be “320 … 320
over 1005” (line 201) along with the fact that the written response provided was correct and the
CLA student stated to the CHA student that “you should tell me what to write” implies that the
CHA student moved to L4 reasoning with regards to theoretical probability in Problem Set 1.
The same though cannot be concluded about the CLA student who seemed to remain ‘confused’
and depended entirely on the CHA student in providing the group’s written response. Note that at
the beginning of the session the IR specified that the CLA student would be the group’s recorder.
Another type of computation that students were required to carry out involved finding the
probability of the union of events (part c of problem 4.9). In this case the CMA student again
moved from L2 to L4 reasoning. The CHA student though exhibited L2 reasoning throughout the
computations involved in part c of the problem and there were no instances of conversation
involving the CLA student during that part of the group work. In particular, the CHA student
stated that “[I]n part c it says ‘or’. So, I will consider first only ‘unfair’ so 600 out of 1005. And
155

then I will consider ‘more than’ which is 500 out of 1005” (lines 160-161). This corresponds to
L2 reasoning since the CHA student used quantitative judgments to compute the probability;
however, this computation was not carried out correctly since the CHA student did not subtract
the intersection of the two events. At the end of the group work another statement made by the
CHA student was that “[S]ince the denominators are the same we don’t need to write them right?
I mean the over 1005. On top it should be 320 here” (lines 219-220). A copy of the student’s
written work revealed that their response to part c was 600 + 320 – 280 = 640. Since the
computation for the probability of the union of events was incorrect the CHA student’s level of
reasoning remained at L2.
On the other hand, the CMA student started off by exhibiting L2 reasoning when
computing the probability of the union of events. Help received by the other group members
along with an examination of class notes helped the CMA student move to L4 reasoning with
regards to this type of computation.
112

CMA: Unfair … 600. More than 50000 … 500. So we should consider the 600
and the 500.

113

OGM: Aren’t we supposed to use 320?

…
121

CMA: Let’s take 320 … then 280.

…
124

CMA: So part c is 780 over 1005? …

126

OGM: Wait guys. Here it is. It’s 600 over 1005. In the class example that she

127
128

showed us before it was red and black. In our case it’s fair and unfair right?
CMA: No, it’s ‘less than’ and ‘more than’.
156

129

OGM: What are you talking about?

130

CMA: Whatever

131

OGM: Write 320 over 500; the other one is 320 over 600. And then subtract …

132

CMA: No way! It’s not out of 600; it’s out of 1005.

133

OGM: Then we subtract 320 over 1005.

134

OGM: We should write 320 over 1005; then 600 over 1005.

135

CMA: No, we only need the 500 and the 600. The 320 involves both
characteristics.

136

OGM: Yes, so we subtract 320 over 1005.

137

CMA: What are you doing? Why are you using 280?

138

OGM: Because it’s about ‘unfair’. We need to take all of those who said ‘unfair’.

139

CMA: Yes, 600.

140

OGM: Ok wait, wait … we should write 600 over 1005

141

CMA: Plus 500 over 1005

At the beginning of Activity 3 all three students monitored in the treatment group could
easily identify impossible events (i.e. sum of two dice being 1, 13, 14, or 15) thus exhibiting L1
reasoning. When asked which sums give the best chance of winning the game that students were
playing in Activity 3 (i.e. which sums are most likely to occur when rolling two dice) the THA
student indicated that “[N]umber 7. Which is six times. And we should write the pairs that give
the number 7: 1-6, 6-1, 5-2, 2-5 … 3-4, 4-3” (lines 281-284). This indicates that the THA student
“uses valid quantitative reasoning to explain comparisons” but did not assign numerical values to
probabilities, thus exhibiting L3 reasoning with regards to theoretical probability. The THA
student then asked the IR “on the table are we supposed to write the pairs? The results?”. At that
157

time the IR responded that students should write down the probability associated with each sum.
The THA student then went ahead to divide the number of outcomes associated with each sum
by 36 thus assigning a correct probability to each sum (L4 reasoning).
In Activity 3 the TMA student and the TLA student moved across three levels of
reasoning (L1  L2  L4). As previously mentioned, both students were able to identify
impossible events at the beginning of the activity thus exhibiting L1 reasoning with regards to
theoretical probability. Later, when asked to compute the probability of each sum when two dice
are rolled, the TMA student correctly started enumerating the outcomes associated with each
sum but used an incorrect denominator when assigning a probability to each sum (L2 reasoning).
That is, instead of dividing by 36, the TMA student divided by 11 (the number of possible sums
when two dice are rolled). With the aid of another group member, the TMA student soon realized
that the denominator should be 36 thus assigning the correct probability to each sum and
indicated that a sum of 7 gave the best chance of winning the game (L4 reasoning). The TLA
student seemed to need more help that the TMA student did from other group members in
enlisting the outcomes associated with each sum (lines 309-360). At first, the TLA student listed
the number of possible outcomes associated with each sum instead of assigning a probability to
each sum (L2 reasoning) but with the aid of another group member realized that each of these
numbers should have been divided by 36. At the end of the activity, the TLA student correctly
indicated that the sum of the probabilities of dice sums should be 36/36 and that a sum of 7 gives
the best chance of winning the game (L4 reasoning).
In Problem Set 3 the three students monitored in the control group worked on two
problems on discrete probability distributions (5.3a and 5.4; See Appendix C). Problem 5.4
required students to construct the probability distribution associated with each of three different
158

scenarios of a game. Students had to use basic probability concepts such as sample space and
theoretical probability of the sum of two dice. Transcripts revealed no instances in which the
CLA student reasoned about theoretical probability. The CHA student and the CMA student both
moved from L2 reasoning to L4 reasoning with regards to theoretical probability.
65

IR: In the first method the player wins if the sum is less than 7.

66

OGM: So, 2 up to 6. So, the probability has to do with getting one of the numbers

67

from 2 to 6. That is, five numbers, So, five out of

68

CHA: 12. Or is it out of 6 since there are six numbers on a die?

69

IR: How many dice are involved here?

70

CHA: Two. So, there are 36 probabilities. So, here we have 5 out of 36?

71

IR: We have two dice. What outcome will give us a sum of 2? What needs to

72

show up on the dice?

73

CHA: 1-1

74

IR: Anything else?

75

CHA: No.

76

IR: So, how many outcomes give a sum of 2?

77

CHA: One

78

IR: How many outcomes give us a sum of 3?

79

CHA: Three

80

IR: Which three?

81

CHA: 1-2, 2-1 and 3-0; ooops there is no 0! So, only two. To get a sum of four we

82

have 2-2, 1-3, 3-1. So, we need to find all the possible results associated
with sums 2 to 6.
159

As the above excerpt indicates, the CHA student was initially confused; the student attempted to
use quantitative reasoning to deal with the situation at hand but the probabilities this student
provided were incorrect (L2 reasoning). Once the IR started asking questions, the CHA student
started to provide correct responses and realized that in order to be able to solve the problem, the
group needed to find the probability associated with each sum that was of interest in the
particular situation (i.e. sums 2 to 6). The CHA student then went ahead to correctly list the
outcomes associated with each of these sums and correctly divided the number of outcomes by
36 to arrive at the probability of each sum (L4 reasoning; lines 71-133).
In Problem Set 3 the CMA student exhibited L2 reasoning at first. When asked about the
probability of getting a sum of 2 when two dice are rolled, he stated “2 over 36 … because there
are two dice” (lines 51 and 54). The IR reminded the CMA student that what is of interest is the
sum of the dice and the CMA student responded “Oh, the sum! Then 1 over 36 because we can
only get a sum of 2 if we get 1-1 on the dice” (line 56). This statement indicates L4 reasoning
since the student assigned a correct numerical probability to an event. When asked about the
probability of getting a sum of 3, the CMA student once again moved from L2 (line 67) to L4
reasoning (lines 74, 76, and 78).
66

IR: Why 2 over 36?

67

CMA: Well, we have 1-2 and … no, 1 over 36.

68

IR: Why 1 over 36?

69

OGM: There is no other outcome.

70

IR: What if the dice are of different colors?

71

OGM: What does that have to do with the numbers on the dice?

72

IR: What outcome did you mention before that gives a sum of 3?
160

73

OGM: If the first die shows a 1 and the second die a 2.

74

CMA: What about the reverse?

75

IR: What do you mean?

76

CMA: So, 2-1 as well.

77

OGM: Oh, I see.

78

CMA: So, the probability for a sum of 3 is 2 over 36.

At first, the CMA student did not consider alternating faces thus providing an incorrect
probability for the sum of 3 when two dice are rolled (L2 reasoning). Later, the CMA student
went on to consider these and provide the correct probabilities associated with each sum (L4
reasoning).
As evident in Table 6.5, students in both the treatment and control groups mostly moved
from a lower to a higher level reasoning with regards to theoretical probability. In some cases,
students in the treatment group performed a higher number of transitions between levels of
reasoning compared to students in the control group and in other cases they performed a smaller
number of transitions.

161

Table 6.5
Levels of Reasoning Exhibited by Students Relative to Theoretical Probability

Activity
1

2

3

4

Treatment
THA
TMA TLA
L2
L2
L2
↓
↓
L4
L4

L2
↓
L4
L1
↓
L3
↓
L4

L4

L1
↓
L2
↓
L4

L2

CHA
L1
↓
L4
↓
L2

2

L4

3

L2
↓
L4

L4
↓
L2
L1
↓
L2
↓
L4

L4

Problem Set
1

Control
CMA
L1
↓
L4
↓
L2
↓
L4

CLA
L1

L4

L2
↓
L4
↓
L2
↓
L4

6.3.3 Experimental Probability of an Event
The activities used in the treatment group provided direct opportunities for students to
express their reasoning regarding experimental probability. However, the problems in the course
textbook on which students in the control group worked in groups, did not directly lend
themselves to discussions regarding experimental probability. Transcripts of student
conversations revealed no instances during which students in the control group reasoned about
experimental probability.
In the treatment group, the three students whose conversations were qualitatively
analyzed exhibited various levels of reasoning relative to experimental probability. These
reasoning levels varied within an activity as well as between activities. As evident from Table
162

6.6, the THA student and the TMA student followed a similar pattern of reasoning in Activity 1,
moving from a lower level reasoning (L1 and L2 respectively) to L4 reasoning. Activity 1 began
by asking students to indicate the sum they believe is most likely to occur when two dice are
rolled. In response to this first question, the THA student reasoned as follows:
26

THA: I believe the answer is 6.

27

OGM: No.

28

THA: Why?

29

OGM: One, two, three … three. Four …

30

THA: Because it is also halfway through 2 and 12

31

OGM starts rolling the dice.

32

OGM: I want to see it in action.

33

THA: What you are doing is a different thing. You simply need to say which

33-34

one shows up more times.

35

OGM: Ok then let’s put 6.

36

THA: All of the pairs have the same chances of coming up. It is not that the

36-37

more times you throw them …

This excerpt shows that the THA student “indicates little or no awareness of any relationship
between experimental and theoretical probabilities” (Jones et al., 1999, p. 15), thus pointing
towards L1 reasoning. Later on in Activity 1 students were asked to determine how close their
group’s experimental probability for each dice sum was to the theoretical probability and then,
whether their group’s experimental probabilities or the whole class experimental probabilities
were closer to the theoretical probability of each dice sum. The THA student indicated that their

163

group’s experimental probabilities were “relatively close” to the theoretical probabilities (lines
309-312) and later reasoned that:
418

THA: Columns 4 and 5 are almost the same. These are closer than these …

418-419

compared to ours these are closer.

420

OGM: Yes, yes

421

THA: So, here where it asks which set of relative frequencies those from your

421-422

group or those from the whole class are closer to the theoretical we will put

423

THA and OGM: the whole class results

424

OGM: are most close

425

THA: are closer to the theoretical probabilities

This excerpt indicates that the THA student “recognizes that the experimental probability
determined from a large sample of trials approximates the theoretical probability” (Jones et al.,
1999, p. 15) thus exhibiting L4 reasoning with respect to experimental probability. In the same
activity, the TMA student indicated that 7 is the most likely sum when two dice are rolled as a
response to the first activity question. When experimental data came into conflict with this
preconceived notion, the TMA student believed that the roller was to blame and requested that
they switch roles:
47

TMA: You and the number 7 are not in good terms?

48

OGM: Yes. 5 … 3

…
56

OGM: 5 … 9

57

TMA: Dude what are you doing?

…
164

60

TMA: Give them to me to roll them.

61

OGM: You are not a roller. When it says you should be a roller you can roll

62

OGM: 7

63

TMA: 7

64

OGM: See? And you were getting stressed!

65

TMA: Goodness! It takes you soooo long to roll a 7!

…
200

TMA: We said at the beginning that sum of 7 would be coming up more, didn’t

200-201

we? Now, based on our results, what patterns do we see?

201

What predictions do we have now?

202

OGM: 6 and 9

203

TMA: Yeah, 6 was lucky

It is evident that the TMA student reverts to subjective reasoning (i.e. believing that generating a
particular outcome depends on who rolls the dice) when experimental data conflicts with
preconceived notions about probability, thus demonstrating L2 reasoning with respect to
experimental probability. Later, when asked to compare their group’s experimental results to the
theoretical probabilities, the TMA student indicates that these are close (lines 256-274). A
comparison of the group experimental probabilities to the theoretical probabilities and whole
class experimental probabilities to the theoretical probabilities, provided the means for the TMA
student to realize that the experimental probabilities obtained from the whole class data were
closer to the theoretical probabilities than the group results alone (lines 321-348) thus exhibiting
L4 reasoning.

165

Note that in Activity 1, the TMA student and the TLA student were in the same group.
Although the TLA student participated actively in the group work by i) providing very brief
responses, ii) converting fractions into percentages for the purposes of reporting probabilities and
iii) volunteering to construct one of the histograms, the oral responses provided by this student
were not elaborate enough to clearly indicate the level of reasoning carried with respect to
experimental probability.
In Activity 2 the three students in the treatment group initially demonstrated L4 reasoning
with regards to experimental probability however, at a later stage in the activity, the THA student
and the TMA student exhibited L2 reasoning while the TLA student demonstrated L1 reasoning.
At the beginning of Activity 2 the three students were able to assign a numerical value for the
experimental probability of an event using the data collected, thus demonstrating L4 reasoning.
At a later stage in the activity, students were asked to compute the theoretical probability that the
sum is 6 when three dice are rolled. The THA student as well as the TMA student initially used
the experimental data to determine this probability (lines 180-186 and lines 92-94 respectively)
indicating a belief “that any sample should be representative of the parent population” (Jones et
al., 1999, p. 15) which corresponds to L2 reasoning. Both students though later realized that they
should have determined these probabilities theoretically and not using the experimental data.
They then proceeded to compute this probability correctly.
The three students in the treatment group exhibited low-level reasoning (THA: L1; TMA:
L2; and TLA: L1) with regards to experimental probability in Activity 3. Once the game winner
was determined, students were asked why they thought the particular group member won. The
THA student responded that “it was out of luck” (line 157) indicating “little or no awareness of
any relationship between experimental and theoretical probabilities” (Jones et al., 1999, p. 15)
166

which corresponds to L1: Subjective Reasoning. In a similar manner, the TLA student addressed
the winner as “you lucky thing!” (line 187) thus, also exhibiting L1 reasoning regarding
experimental probability. On the other hand, the TLA student indicated that “we should have
placed chips on 7. That’s the one that comes up most often” (line 58). However, the TLA student
made this statement after only nine rolls of the two dice, three of which resulted in a sum of 7.
This indicated that the TLA student “puts too much faith in small samples of experimental data
when determining the most or likely event; believes that any sample should be representative of
the parent population” (Jones et al., 1999, p. 15).
In Activity 4 the THA student and the TMA student followed alternating levels of
reasoning (THA: L4  L1  L4; TMA: L4  L2  L4) while the TLA student went from L2
to L4 reasoning with regards to experimental probability. The THA student was initially able “to
determine a numerical value for the experimental probability” using the data (line 171), thus
demonstrating L4 reasoning. Students were later asked to compare their results with those of
another group and to indicate
209

THA: Why do you think the probabilities differ? Because they are

209

probabilities!

210

IR: Why did you have differences?

211

OGM: Because the numbers are random.

212

THA: So, let’s write that the random numbers are different from the random

212-213

numbers of the other group. Also, write down that our result was 80%

213

whereas the other group got 76%.

The response of the THA student that probabilities differ “[B]ecause they are probabilities!”
“indicates little or no awareness of any relationship between experimental and theoretical
167

probability” thus exhibiting L1 reasoning (Jones et al., 1999, p. 15). However, when another
group member (OGM) related group differences to the concept of randomness, the THA student
agreed and went on to provide an explanation based on this concept. At a later stage the THA
student was able to determine the experimental probability based on the whole class results (lines
246-273) thus exhibiting L4 reasoning. Similar to the THA student, the TMA student was
initially able to provide a numerical value for an experimental probability thus demonstrating L4
reasoning. When asked to determine an experimental probability based on the whole class
results, the TMA student was once more able to provide a numerical value for it thus exhibiting
L4 reasoning:
196

TMA: Based on the results from all groups, what is the probability that the

196-197

player makes at least two baskets in three free-throw attempts? So, we

197-198

should find the mean. 40 plus 38 plus 37 plus 44 plus 36 plus 38

198

divided by 6. Miss can you come over?

…
205
205

IR: The total number of times that this experiment was performed by each
group was what?

206

OGN: 50

207

IR: Where did you include the 50?

208

OGM: What do you mean?

209

TMA: Divide by 50.

210

OGM: And after that, all divided by 6.

211

TMA: Divide by 50.

212

IR: What should be divided by 50?
168

213

TMA: 40 over 50; 38 over 50. And then the result divided by 6.

The low ability student in the treatment group (TLA), first demonstrated L3 reasoning
with respect to experimental probability. In particular, when asked to determine the probability
of making at least two baskets using the experimental data, the TLA student responded “one,
two, three, four, … 44. So, we will write P of at least two baskets is 44” (line 142); that is, the
student provided a frequency value when asked to provide a probability. At a later stage, the
TLA student was able to determine a numerical value for the experimental probability based on
the whole class results thus exhibiting L4 reasoning with regards to experimental probability:
172

OGM: So, do we add everything up and divide by 50?

173

TLA: Don’t we need to add everything up, divide by 6 and then by 50?

174

OGM: Ok. Would you like to write this or shall I write this?

175

IR: What is the total number of outcomes for the whole class?

176

TLA: 300. Oh, so do we take this total and divide by 300?

177

IR: Yes.

178

TLA: It comes out to 233.

179

OGM: And divide that by 300.

180

TLA: Miss, 0.77.

181

OGM: So, 77%.

Table 6.6 indicates the levels of reasoning demonstrated by the three students in the
treatment group relative to the construct of experimental probability. Note that the problem sets
did not provide students in the control group with any direct or indirect opportunities to reason
about experimental probability and so, since there was no data on these students, Table 6.6 does
not include a presentation of levels of reasoning exhibited by them.
169

As can be seen in Table 6.6, the three students monitored in the treatment group moved
from a lower to a higher level reasoning in Activity 1 and Activity 4; moved from a higher to a
lower level reasoning in Activity 2; and remained stable at a low level reasoning in Activity 3.
Table 6.6
Levels of Reasoning Exhibited by Students Relative to Experimental Probability

Activity
1

2

3
4

THA
L1
↓
L4
L4
↓
L2
L1
L4
↓
L1
↓
L4

Treatment
TMA
L2
↓
L4
L4
↓
L2
L2
L4

TLA

L4

L1
L3
↓
L4

6.3.4 Summary
An examination of Tables 6.3, 6.5 and 6.6 regarding students’ reasoning levels relative to
sample space, theoretical probability, and experimental probability respectively, revealed the
following: i) The activities provided students with more direct as well as indirect opportunities to
reason about sample space and experimental probability compared to the problem sets; ii) The
problem sets did not provide any direct or indirect opportunities for students in the control group
to reason about experimental probability and so, there was no data on these students relative to
this construct; iii) Students’ reasoning levels were the most stable with regards to sample space
in comparison to theoretical or experimental probability; iv) Students exhibited higher levels of
reasoning with regards to sample space (mostly L4) compared to theoretical or experimental
170

probability across all of the four activities and problem sets; v) With regards to experimental
probability, the three students monitored in the treatment group moved from a lower to a higher
level reasoning in Activities 1 and 4; moved from a higher to a lower level reasoning in Activity
2; and remained stable at a low level reasoning in Activity 3; and vi) With regards to theoretical
probability students in both the treatment and control groups mostly moved from a lower to a
higher level reasoning.
6.4 Effects of Instructional Treatment on Understanding of Conditional Probability and
Independence
In order to examine the effects of the instructional treatment on students’ understanding
of conditional probability and independence, the audio-taped conversations of the three students
selected in the treatment group and the three students selected in the control group as they
worked in small groups on Activity 2 and Problem Set 2 (see Appendix C) respectively, were
qualitatively analyzed. The six students exhibited the levels of reasoning shown in Tables 6.8
and 6.9 with regards to conditional probability and independence respectively. These two
concepts are represented by two constructs on the Jones et al. (1999) framework i.e. conditional
probability and independence.
6.4.1 Conditional Probability
Specific to conditional probability, both Activity 2 as well as Problem Set 2 provided
students with direct opportunities to reason about conditional probability since both of these
instructional materials focused on this concept. Activity 2 required students to roll a set of three
dice of different colors 50 times, record the outcome and sum of the dice on each roll in a table,
and compute various probabilities. In this activity, the THA and TMA students both
demonstrated L4 reasoning with regards to conditional probability. An examination of the
171

transcripts revealed no conversational instances in which the TLA student expressed reasoning
relative to conditional probability. Once students performed the experiment, they were asked to
use their results to i) compute the relative frequency of getting a sum of 6 when three dice are
rolled and ii) if the sum is 6, to determine the relative frequency that one of the dice results in 3.
31

THA: How many times we got a sum of 6?

…
35

THA: Two

36

THA and OGM: Twice out of 50.

37

THA: Based on your results, if the sum is 6 what is the relative frequency that one

38

of the dice results in 3?

…
40

THA: From these two times that we put as an answer here, did we have a three in
any of them?

41

OGM: Let’s check.

42

THA: Here. So, once.

A copy of the group’s written work revealed that students assigned a correct numerical answer
(i.e. ½ = 50%) for the question that required a conditional probability. So, the THA student
exhibited L4 reasoning relative to conditional probability. The TMA student also demonstrated
L4 reasoning in the same situation.
32
33

TMA: 7 times? Based on your results, what is the relative frequency to get a sum
of 6 when you roll three dice?

34

TMA and TLA: 7 over 50.

35

TMA: Yes. Based on your results, if the sum is 6, what is the relative frequency
172

36
37

that one of the dice results in 3?
TMA: So, if we got a sum of 6, how many of those times did we get a 3?

…
39

OGM: If we had a 3, how many of those times did we have a sum of 6?

40

TMA: If one of the dice was 3 … wait! If the sum was 6…

…
43

TMA: One…

44

OGM and TMA: Two, three

45

TMA: Four … 4 out of 7.

A copy of this group’s written work indicated that they provided a correct value for the relative
frequency required (i.e. 4/7 = 57%). So, the TMA student demonstrated L4 reasoning with
regards to conditional probability.
Later on in Activity 2, students were asked “to find the theoretical probability that, when
three dice are rolled, one of the dice results in 3 given that the sum is 6” (see Appendix C) by
responding to a set of questions. These questions acted as a structured set of steps that aimed to
help students arrive at the required theoretical conditional probability. Both the THA student and
the TMA student correctly identified that there are 216 possible outcomes when three dice are
rolled (line 89 and line 112 in the respective transcripts) and correctly indicated that the
probability to get a sum of 6 when three dice are rolled is 10/216 by considering the outcomes
that result in a sum of 6 (lines 96-112, 173 and lines 82-91, 151 in the respective transcripts).
Next, both students correctly indicated that the probability of getting a sum of 6 (i.e. 10/216)
should be used in the denominator of the required conditional probability and that the numerator
(i.e. probability that one of the dice results in 3 and the sum is 6) should be 6/216.
173

188
189
190

THA: …The probability to get a 3 with a sum of 6 is 6 out of 216. The probability
that the sum is 6…
OGM: One sixth

…
204

OGM: But we said it is 10.

205

THA: 10 over 216. Hmmm, so, 6 over 216

206

OGM: 0,03

207

OGM: And the probability for the sum to be 6

208

THA: We found that … 0,05

Similarly the TMA student reasoned as follows:
146

TMA: So, write here, one of the dice is 3. And there, sum is 6. Over sum is 6.

147

OGM: One of the dice is 3 … 3 to the power of 3.

148

TLA: 3 to the power of 3 is 27.

149

TMA: The probability to get a sum of 6 is this over here.

150

OGM: 27 out of 216. Isn’t it?

151

TMA: No, 10 over 216.

…
161

IR: How many outcomes resulted in a sum of 6 and contained a 3?

162

OGM: 6

163

OGM: Six out of?

164

OGM: 10. So, the numerator is 6 over 10.

…
171

TMA: Hey, it should be 6 over 216 here, not over 10.
174

Since both the THA and TMA students were able to assign these numerical probabilities the
level of reasoning of reasoning exhibited by these students with regards to conditional
probability was L4.
In Problem Set 2 the three students monitored in the control group worked on problem
4.23 from the course textbook which made use of the frequency contingency table provided in
Table 6.7 and asked for the following conditional probabilities:
Table 6.7
Problem 4.23 (Levine, Krehbiel, and Berenson, 2010)

U.S. Tax Code
Fair
Unfair
Total

Income Level
Less than $50,000 More than $50,000
225
180
280
320
505
500

Total
405
600
1,005

a. Given that a respondent earns less than $50,000, what is the probability that he or
she said that the tax code is fair?
b. Given that a respondent earns more than $50,000, what is the probability that he
or she said that the tax code is fair?
All three students monitored in the control group ended up providing correct numerical
probabilities for the two conditional probabilities required in problem 4.23. However, initially
they all exhibited L2 reasoning since they all provided an incorrect numerical response to part a.
In particular, the CHA and the CMA students ignored the conditioning event and stated that the
answer to part a was 225/1005 (CHA: lines 7 and 44-46; CMA: line 10) while the CLA student
claimed that the answer to part a was 225/405 (line 34). Following this initial statements though,
all students went on to reason at L4 and ended up assigning correct numerical probabilities for
both parts a and b of problem 4.23.
59

CHA: The question says to look at those who earn less than 50000 and then to
175

60

look for the tax code being fair.

61

OGM: What did we put after the vertical line?

62

CMA: Less than 50000.

63

CHA: So, being fair will be placed in the denominator.

64

CMA: Less than 50000 goes in the denominator and the other one on top. Both of

65

these together go on the numerator.

…
69

CMA: Less than 50000 is 505 over 1005. This goes in the denominator.

70

CHA: Ok. And on top? 225 over 1005.

The CLA student reasoned as follows in the first part of problem 4.23.
34

CLA: So, 225 over 405

35

OGM: What did you do there?

36

CLA: I divided by the total. What we did with ‘give that’, shouldn’t it provide

37

some other piece of information about something that happened first?

38

OGM: Wait.

39

CLA: Here there are those who believe it is fair and those who think it is unfair.

40

OGM: No, there are those who earn less than 50000 or more than 50000.

41

IR: How many people are there who earn less than 50000?

42

CLA: 225

43

OGM: 505

44

CLA: Oooh, 505

45

IR: Out of those, how many think the tax code is fair?

46

CLA: So, 225 over 505.
176

Lines 36-37 indicate that although the CLA student recognized that the probability of events
change when we deal with conditional probability, this “recognition is incomplete and is …
restricted to events that have previously occurred” (Jones et al., 1999, p. 15). In this manner, the
CLA student demonstrated L2 reasoning with respect to conditional probability. However,
through communication with other members of the group and the IR, the CLA student ended up
assigning a correct numerical probability to part a of the problem, and later to part b as well.
Table 6.8 provides a summary of the reasoning levels exhibited by the six students
monitored in the treatment and control groups relative to conditional probability.
Table 6.8
Levels of Reasoning Exhibited by Students Relative to Conditional Probability

Activity
2

Treatment
THA TMA TLA
L4
L4

Problem Set
2

Control
CHA CMA
L2
L2
↓
↓
L4
L4

CLA
L2
↓
L4

6.4.2 Independence
Activity 2 and Problem Set 2 provided students with direct opportunities to reason about
independence. Activity 2 required students to determine whether getting a sum of 6 and one of
the dice resulting in 3 when three dice are rolled were independent. Students were asked to
indicate this twice during the activity; first following the performance of an experiment with
three dice and later, after computing the theoretical probability of rolling a 3 on one of the three
dice given that the sum was 6. Following the completion of the experiment, the THA student said
that
57-58 THA: If one of them is 3 then the other two dice must be 1-2 or vice versa in
177

order to get a sum of 6.
59

IR: Ok. If the answers to 2a and 2b were the same what would that mean?

60

THA: That they don’t affect each other. So, in this case, they affect each other.

This excerpt indicates that the THA student exhibited L3 reasoning with regards to the concept
of independence since the student could “differentiate independent and dependent events” but at
that time had not reached L4 reasoning which involved using numerical probabilities to
distinguish such events. In contrast, the TMA student stated that since the answers to 2a (i.e.
Based on your results, what is the relative frequency that you will get a sum of 6 when you roll
three dice) and 2b (i.e. Based on your results, if the sum is 6, what is the relative frequency that
one of the dice results in 3?) were not the same “they would not be affected. I mean they are not
the same thing”, implying that the events were independent (lines 48-49). The reasoning
provided by the TMA student indicated that this student could not differentiate independent and
dependent events and pointed towards L1 reasoning.
At the end of Activity 2, students were once more asked to determine whether getting a
sum of 6 and one of the dice resulting in 3 were independent. The THA student was able to use
numerical probabilities to show that the two events were dependent. Although one of the
probabilities computed was incorrect (i.e. probability that one of the dice results in 3), the steps
followed in identifying whether the events were independent as well as the remaining
computations, and conclusion were correct. Therefore, the THA student ended up exhibiting L4
reasoning relative to the concept of independence. In contrast, the TMA student, along with the
other members of his group, found this task to be too challenging and gave up without asking for
help from the IR. So, the TMA student remained at L1 reasoning with respect to the concept of
independence.
178

In problem 4.23 (Table 6.7) the three students monitored in the control group were asked
to indicate whether income level was independent of attitude about whether the tax code was
fair. The three students checked whether P(A|B) = P(A) in order to determine if the events were
independent. All three students were able to use numerical probabilities correctly to conclude
that the events were dependent thus exhibiting L4 reasoning with regards to the concept of
independence.
Table 6.9 provides a summary of the reasoning levels exhibited by the six students
monitored in the two groups with regards to the concept of independence.
Table 6.9
Levels of Reasoning Exhibited by Students Relative to Independence

Activity
2

Treatment
THA TMA TLA
L3
L1
↓
L4

Problem Set
2

Control
CHA CMA
L4
L4

CLA
L4

6.4.3 Summary
Both Activity 2 and Problem Set 2 provided students with direct opportunities to reason
about conditional probability and independence. Moreover, both of these instructional materials
afforded opportunities in which students needed to draw on previous knowledge of basic
probability concepts (sample space, theoretical probability of simple and joint events,
experimental probability) in order to compute conditional probabilities and decide whether
events were independent.
Overall, i) students’ reasoning with regards to conditional probability and independence
was stable; ii) students exhibited high levels of reasoning with regards to both concepts; iii) in
179

the case of conditional probability the TLA and TMA students demonstrated L4 reasoning
whereas in the case of independence, all three students monitored in the control group exhibited
L4 reasoning; and iv) with regards to conditional probability all three students monitored in the
control group moved from L2 to L4 reasoning.
6.5 Effects of Instructional Treatment on Understanding of Discrete Probability
Distributions
In order to examine the effects of the instructional treatment on students’ understanding
of discrete probability distributions, the audio-taped conversations of the six students selected in
the treatment and control groups as they worked in small groups, on Activity 3 (see Appendix C)
and Problem Set 3 (see Appendix C) respectively, were examined. Note that the framework by
Jones et al. (1999) did not specifically include discrete probability distributions as a construct.
However, in both of these sets of instructional materials students needed to make connections to
basic probability concepts; reasoning on such concepts was accounted for in the framework.
Activity 3 involved playing a game in which students rolled a pair of dice and each time
removed a chip they might have placed over the number corresponding to the sum of the two
dice. The winner was the person who had all of his/her chips removed first. Therefore, when
distributing their chips over the number line on which the numbers 1 through 15 were written
and with the aim of winning in mind, it would have been ideal if students made connections to
the theoretical probability associated with each sum when two dice are rolled i.e. the probability
distribution of rolling two dice.
In Activity 3, the term ‘discrete probability distribution’ was only mentioned in the title
and in the last part of the activity. Throughout the rest of the activity students dealt with other
basic probability concepts when responding to questions posed and at times, they dealt only
180

indirectly with discrete probability distributions. For example, once students placed the 12 chips
given to each of them on the number line, they were asked to indicate how many chips they had
placed above each number and provide an explanation for the way they had placed their chips.
This aimed to provide students with an opportunity to indirectly deal with discrete probability
distributions. Table 6.10 indicates how students distributed their chips on the number line along
with the explanation provided by each as to the way they distributed their chips. This information
was available on a copy of students’ written work. Notice that, in their written explanations, only
the TMA student made connections to the theoretical probability associated with each sum when
deciding how to place the dice on the number line. The TMA student predicted the most likely
events using informal quantitative judgments (i.e. did not assign numerical probabilities to each
sum) however, he reverted to subjective judgments once the group started rolling the dice by
expressing a belief that the outcome depends on who rolls the dice. Thus, he demonstrated L2
reasoning with regards to theoretical probability. The TLA student also used subjective
reasoning (L2) with regards to theoretical probability when placing chips on the number line (i.e.
the student placed chips on even numbers because “I like even numbers”).

181

Table 6.10
Activity 3: Distribution of Chips By Students and Explanation
Number on the
Number Line
(Dice Sum)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Explanation

THA

“I placed
randomly.”

1

2
2
2
2
1

1

1

my

3

1

3
1
1
1
1

TLA

1

2
3

TMA

2

4
1

chips “More chips towards “I placed my chips on
the center (number 6-9). even numbers because I
Numbers have higher like even numbers.”
probabilities.”

At the end of Activity 3 students in the treatment group dealt directly with the probability
distribution for the sum of two dice. Students were provided with a table on which the first
column was a list of the possible sums of rolling two dice and were required to fill in the second
column by providing the numerical probability corresponding to each sum. Next, they were
asked to sum up the probabilities with the aim of helping them realize that the probabilities
should sum up to 1. All three students monitored in the treatment group were able to compute the
theoretical probability associated with each sum by considering the number of outcomes that
give rise to each sum and dividing this number by 36. Moreover, all groups indicated that the
sum of the probabilities should be 36/36 = 1. In the case of the THA and TLA students, initially
182

the probabilities provided by their corresponding groups did not sum up to 1. These students
received this as an indication that they must have made a mistake on the table and went back to
check and correct the probabilities provided for the sums of two dice. In this manner, all three
students exhibited L4 reasoning with regards to theoretical probability (and indirectly, with
regards to discrete probability distributions).
With regards to Problem Set 3, all three students monitored in the control group were
able to compute the expected value of a discrete random variable with ease when a table listing
the possible outcomes and associated probabilities was provided to them (problem 5.3; see
Appendix C). However, all three students found problem 5.4 quite challenging. In this problem,
students were required to construct the probability distribution for a discrete random variable
themselves. The problem involved a game scenario on which three types of bets could be placed.
All three students realized that they should create a table on which one column should list the
outcomes associated with the bet being placed and a second column should indicate the
probability associated with each outcome. The CHA student said that
38
39

CHA: I think we are supposed to create a table here with two columns and carry
out the multiplications like we did in the previous problem.

while the CMA student indicated that
34

CMA: Ok so we have three methods of playing the game. In the first one we win

35

if 2 through 6 come up. We need to find the probability associated with

36

each game method. We need to have a table like this one.

and the CLA student realized that
28

CLA: We will have to create the table ourselves here.

…
183

33

CLA: Well, there should be ‘under’ and ‘over’ on the table. The player wins or

34

loses. There is a bet. So, we should have ‘under 7’ so let’s write the values
below 7 and ‘over 7’.

Following these statements, both the CHA and CMA students led their groups in constructing the
probability distribution for the sum of two dice by listing the outcomes associated with each sum
and dividing the number of outcomes by 36. In this manner they demonstrated L4 reasoning with
respect to theoretical probability. Moreover, they labeled outcomes as ‘win’ or ‘lose’ and
correctly computed the probability of winning or losing under each of the three game scenarios.
On the other hand, although the CLA student realized that a table should be constructed which
should accommodate for the various outcomes of interest, the student then stated that “I don’t
understand what the problem is asking for” (line 63). The other members of the group stated that
they should list the outcomes that give rise to each sum; the CLA student participated in this
process however, it was evident that the student called out these outcomes without understanding
why this process had to be carried out (line 89: “I don’t understand this at all”). The group ended
up providing only the frequencies associated with each sum and did not go on to convert these
into probabilities.
6.6 Effects of Instructional Treatment on Understanding of Binomial Distribution
In order to examine the effects of the instructional treatment on students’ understanding
of the binomial distribution, the audio-taped conversations of the six students monitored in the
treatment and control groups as they worked in small groups, on Activity 4 (see Appendix C)
and Problem Set 4 (see Appendix C) respectively, were examined. Note that the framework by
Jones et al. (1999) did not specifically include binomial distribution as a construct. However, in

184

both of these instructional materials students needed to make connections to basic probability
concepts; reasoning on such concepts is accounted for in the framework.
In Activity 4 students were asked to compute the probability of making at least two
baskets in three free-throw attempts, in the case of a basketball player who has a constant
probability of making the basket P(B) of 70% and a constant probability of not making the
basket, P(N), of 30%. Students were asked to compute this probability in three different ways: i)
Construct the sample space for three free-throw attempts, compute the probability of each
outcome in the sample space, and then use these individual probabilities and the addition law of
probability to compute the probability of making at least two baskets; ii) use a table of random
numbers to indicate data on 50 sets of three free-throw attempts and use these data to compute
the experimental probability of making at least two baskets; and iii) use the formula for the
binomial distribution to compute the probability of making at least two baskets.
With regards to method i) described above, the three students monitored in the treatment
group were able to correctly construct the sample space for three free-throw attempts thus
exhibiting L4 reasoning with regards to sample space. Next, the THA and the TLA students
demonstrated L4 reasoning with regards to theoretical probability (and subsequently with regards
to the binomial distribution) since they were able to assign correct numerical probabilities to
each outcome in the sample space and then, correctly computed the probability of making at least
two baskets. The TMA student demonstrated L2 reasoning with respect to theoretical probability
since the student used quantitative judgments to compute the probability of each outcome but did
so incorrectly:
56

TMA: Do we do 70 plus 70 plus 30 divided by 3? Times 100?

185

Relative to method ii) the THA and the TMA students were able to use the table of random
numbers to calculate the experimental probability of making at least two baskets thus exhibiting
L4 reasoning with regards to experimental probability (and subsequently with regards to the
binomial distribution). The TLA student was also able to use the table of random numbers but
instead of assigning a numerical probability the student provided a frequency value for the
experimental probability of making at least two baskets; thus, the TLA student demonstrated L3
reasoning with respect to experimental probability (and subsequently with respect to the
binomial distribution). In the case of method iii) in which students were asked to use the general
formula for computing a probability from the binomial distribution, a copy of each group’s
written work indicated that students were able to correctly compute the required probabilities
from the binomial distribution (Question 6, Activity 4; See Appendix C). However, transcripts of
student conversations revealed that all three students monitored in the treatment group had
difficulties applying the general formula in the context of the activity.
In Problem Set 4 the three students monitored in the control group dealt directly with the
binomial distribution by working on problem 5.13 in the course textbook. All three students were
able to define n (the number of observations in the sample space), X (the number of events of
interest), and p (the probability of an event of interest) as well as to compute the mean and
standard deviation of the binomial distribution with ease. Moreover, all three students were able
to apply the general formula for computing the probability from the binomial distribution and
assigned correct numerical probabilities as responses to parts b and c of problem 5.13.
6.7 How Results Address The Research Questions
Recall that the research questions posed in this study aimed to examine the effects of two
instructional methods on college students’ achievement and understanding of probability.
186

Instructional Method A was used in the control group and involved the use of lectures and smallgroup cooperative learning sessions during which students solved probability problems.
Instructional Method B was used in the treatment group and involved lectures and small-group
cooperative learning sessions during which students used activities involving probability
experiments that generate real data to make connections between experimental and theoretical
probability.
The tables constructed in this chapter demonstrate the levels of reasoning that the six
students monitored in the control and treatment groups exhibited relative to the concepts of
sample space (Table 6.3), theoretical probability (Table 6.5), experimental probability (Table
6.6), conditional probability (Table 6.8) and independence (Table 6.9). The aim of constructing
these tables was to be able to compare the levels of reasoning exhibited by students in the control
and treatment groups in the attempt to identify which of the two instructional methods had a
better effect on students’ understanding of probability.
As evident from Tables 6.3 and 6.6, the activities used in the treatment group provided
students with more direct as well as indirect opportunities to reason about sample space and
experimental probability compared to the problem sets used in the control group. In particular,
the problem sets did not provide any direct or indirect opportunities for students in the control
group to reason about experimental probability. Given these, it is not surprising that, as specified
in Chapter 5, students in the treatment group performed better on the multiple-choice items
relating to the Law of Large Numbers (Item 6 on the pre-test and post-test) and the concept of
sample space (Items 7 and 11 on the pre-test and post-test) than students in the control group.
Specific to Item 6 the percent-correct responses decreased by 11.7% in the control group and by
3.7% in the treatment group from pre-test to post-test. On Item 7 the percent-correct responses
187

increased by 5.8% in the case of the control group and by 18.6% in the treatment group from pretest to post-test. Last, on item 11, the percent-correct responses decreased by a substantial 58.8%
in the case of the control group and increased by 3.7% in the treatment group. Considering these,
Instructional Method B (treatment) produced better results than Instructional Method A relative
to experimental probability and the concept of sample space.
An examination of Tables 6.5, 6.8 and 6.9 which demonstrate the levels of reasoning
exhibited by students relative to theoretical probability, conditional probability and independence
respectively, indicate that students in both the control and treatment groups were provided with
opportunities to reason about these concepts. As evident from Table 6.5, i) students in both the
control and treatment groups mostly transitioned from a lower to a higher level reasoning with
regards to theoretical probability; ii) in some cases students in the control or treatment group
exhibited a stable level of reasoning; and iii) students in the control group performed 0-3
transitions between levels of reasoning while students in the treatment group performed 0-2
transitions between levels of reasoning relative to theoretical probability. The results presented in
Tables 6.8 and 6.9 reveal that students in the control and treatment groups either remained at L4
level of reasoning or performed one transition from a lower to a higher level reasoning regarding
conditional probability and independence. Based on these results, it is difficult to identify
whether one group performed better than the other relative to these concepts and so, which
instructional method had a better effect on students’ understanding of theoretical probability,
conditional probability and independence.
6.8 Summary
This chapter presents the results of the qualitative analysis of the data collected through
audio-taped student conversations of the six students monitored in the treatment and control
188

groups. Students’ reasoning relative to basic probability concepts (sample space, experimental
probability of an event and theoretical probability of an event), as well as relative to conditional
probability and independence was evaluated using the Jones et al. (1999) framework in a direct
manner since the framework accounted for these concepts. However, the framework does not
accommodate for the concepts of discrete probability distributions and binomial distribution
which were also examined in this study. So, students’ level of reasoning exhibited in excerpts
associated with discrete probability distributions (in Activity 3/Problem Set 3) and the binomial
distribution (in Activity 4 /Problem Set 4) could not be directly coded using the framework. In
Activity 3/Problem Set 3 and in Activity 4/Problem Set 4 students needed to make connections
to basic probability concepts; reasoning on such concepts is discussed in the framework.
Excerpts in which students made connections to basic probability concepts were analyzed using
the Jones et al. (1999) framework. Any excerpts relative to the concepts of discrete probability
distributions and the binomial distribution in Activity 3/Problem Set 3 and in Activity 4/Problem
Set 4 respectively, in which students did not make connections to basic probability concepts
were discussed in separate sections (sections 6.4 and 6.5 of this chapter).
With regards to students’ reasoning levels relative to basic probability concepts, these were
the most stable (i.e. involved fewer transitions between levels of reasoning) in the case of the
concept of sample space compared to theoretical or experimental probability. Students exhibited
higher levels of reasoning with regards to sample space (mostly L4) compared to theoretical or
experimental probability across all of the four activities and problem sets. With regards to
experimental probability, the three students monitored in the treatment group moved from a
lower to a higher level reasoning in Activities 1 and 4; moved from a higher to a lower level
reasoning in Activity 2; and remained stable at a low level reasoning in Activity 3. With regards
189

to theoretical probability students in both the treatment and control groups mostly moved from a
lower to a higher level reasoning. In some cases, students in the treatment group performed a
higher number of transitions between levels of reasoning relative to theoretical probability
compared to students in the control group and in other cases they performed a smaller number of
transitions.
Relative to conditional probability and independence, students’ reasoning was stable and
students exhibited high levels of reasoning with regards to both concepts. In the case of
conditional probability the TLA and TMA students demonstrated L4 reasoning whereas in the
case of independence, all three students monitored in the control group exhibited L4 reasoning.
With regards to conditional probability all three students monitored in the control group moved
from L2 to L4 reasoning.
In the case of discrete probability distributions, the three students monitored in the
treatment group were able to make connections to theoretical probability when computing the
probability distribution of a discrete random variable. These students were able to compute the
theoretical probability associated with the each sum when rolling two dice. Overall, the three
students in the treatment group exhibited L4 reasoning with regards to theoretical probability
(and indirectly, with regards to discrete probability distributions). In the control group, all three
students were able to compute the expected value of a discrete random variable with ease when a
table listing the possible outcomes and associated probabilities was provided to them. However,
when asked to construct the probability distribution for a random variable themselves, all three
students in the control group found such a task to be challenging.
The last concept examined in this study was that of binomial distribution. The three
students in the treatment group were able to construct the sample space for a random variable
190

that follows the binomial distribution thus exhibiting L4 reasoning with regards to sample space.
The THA and the TLA students demonstrated L4 reasoning with regards to theoretical
probability (and subsequently with regards to the binomial distribution) since they were able to
assign correct numerical probabilities to each outcome in the sample space and then, correctly
computed the probability of an event. The TMA student demonstrated L2 reasoning with respect
to theoretical probability since the student used quantitative judgments to compute the
probability of each outcome but did so incorrectly. In addition, the THA and the TMA students
were able to use a table of random numbers to calculate the experimental probability of an event
thus exhibiting L4 reasoning with regards to experimental probability (and subsequently with
regards to the binomial distribution). The TLA student was also able to use the table of random
numbers but instead of assigning a numerical probability the student provided a frequency value
for the experimental probability of the event; thus, the TLA student demonstrated L3 reasoning
with respect to experimental probability (and subsequently with respect to the binomial
distribution). When it came to using the general formula for computing a probability from the
binomial distribution, a copy of each group’s written work indicated that students were able to
correctly compute the required probabilities from the binomial distribution. However, transcripts
of student conversations revealed that all three students monitored in the treatment group had
difficulties applying the general formula in the given context. The three students monitored in
the control group were able to define n (the number of observations in the sample space), X (the
number of events of interest), and p (the probability of an event of interest) as well as to compute
the mean and standard deviation of the binomial distribution with ease. Moreover, all three
students were able to apply the general formula for computing the probability from the binomial
distribution and assigned correct numerical probabilities as responses to textbook problems.
191

Chapter 7 provides a discussion of the quantitative and qualitative results presented in
Chapters 5 and 6 respectively, including possible explanations regarding students’ performance.
In addition, Chapter 7 identifies the conclusions drawn from this study along with implications
for further research.

192

CHAPTER 7
SUMMARY, DISCUSSION AND RECOMMENDATIONS
This study was designed to examine the role of tasks on college students’ achievement and
understanding of probability. This chapter is organized into five sections: i) summary of the
study; ii) summary of the findings; iii) discussion of findings; iv) strengths and limitations; and
v) implications and recommendations for further research.
7.1 Summary of the Study
Since the late 1950s there has been a strong call for an increase in the inclusion of probability
in mathematics curricula in the USA as well as in Europe (Exarchakos, 1988; Lordou-Kaspari,
2003; Jones, 1970; MAA, 1998; NCEE, 1983; NCSM, 1977; NCTM, 2000). Over the past
couple of decades, there have been various reform initiatives concerning the content and means
of instruction in mathematics classrooms at the college level with recommendations set forth that
lectures be replaced by more active learning methods in which group work is used (MAA, 1998).
Specific to the area of statistics and probability, many researchers have recommended that there
be a change in the way statistics courses are taught (Chance, 1997; Garfield, 1994; Shaughnessy,
1981; see also Keeler & Steinhorst, 2001). “One area that has received less focus in this literature
is the teaching of probability” (Keeler & Steinhorst, 2001, Retrieved April 24, 2009 from
http://www.amstat.org/publications/jse/v9n3/keeler.html). Much has been written about people’s
misconceptions and use of heuristics regarding judgment under uncertainty but there has been a
lack of research on solutions to this phenomenon (Keeler & Steinhorst, 2001).

193

7.1.1 Purpose
With the above issues under consideration, this study examined the role of particular tasks on
college students’ achievement and understanding of probability. The research questions
addressed in this study involved the use of two instructional methods:
Instructional Method A: Using lectures and small-group cooperative learning sessions during
which students solve probability problems and
Instructional Method B: Using lectures and small-group cooperative learning sessions during
which students use activities involving probability experiments that generate real data to make
connections between experimental and theoretical probability.
Given Instructional Method A and Instructional Method B, the research questions addressed
were:
3) What are the effects of using each of these instructional methods on college students’
achievement on probability and on their understanding of experimental and theoretical
probability?
4) Does Instructional Method B have a better effect on college students’ achievement on
probability and on their understanding of experimental and theoretical probability than
Instructional Method A?
7.1.2 Methods
A mixed methods design was used to address the research questions for this study. The
design included treatment and control groups, each comprised of students in three sections of an
introductory statistics course taught by the researcher in spring 2010 at a college in Cyprus.
Formal instruction on probability occurred during the second half of the semester and lasted for
seven weeks. A pre-test comprising of 14 multiple-choice items and a post-test comprising of 15
194

multiple-choice items and 2 open-ended items were utilized. During the study, students in the
treatment group worked in small groups on four in-class activities about experimental and
theoretical probability, and students in the control group worked in small groups on solutions to
four sets of probability problems. Three students in the treatment group and three students in the
control group were monitored by having their conversations audio-recorded as they worked in
groups.
Participants included 44 students across the three sections of the introductory statistics
course. Students were not randomly assigned to these sections but instead were placed in them
by the course coordinator in consultation with the academic board at the college based on
students’ English language proficiency. Two of the sections were combined to form the
treatment group which was taught using Instructional Method B; one section included students of
moderate/high English language proficiency while the other included students of all levels of
English language proficiency. The third group acted as the control group and was taught using
Instructional Method A; students in this group were of low English language proficiency. Due to
this type of placement it was challenging to disentangle the extent to which treatment effects
reflected in the results of the study were due to English language proficiency or the treatment
itself.
In order to investigate whether the two instructional methods under consideration had a
significant effect on students’ achievement and understanding of probability, first an analysis
was carried out to identify whether the students in the three course sections had comparable
initial probability knowledge. The results of the analysis indicated that the students in the control
group had comparable initial probability knowledge to the students in the treatment group.

195

Quantitative as well as qualitative analysis were carried out to address the research questions.
First, a comparison of gain scores within each group from pre-test to post-test was performed in
order to determine the effects of each instructional method on students’ achievement in
probability. That is, the pre-test scores of the treatment group were compared to the post-test
scores of the same group in order to determine if the instructional treatment had an effect on
students’ achievement. Also, the control group’s pre-test and post-test scores were compared to
determine if the instructional method used in this group had an effect on students’ achievement.
Second, a comparison of normalized gain scores was performed in order to establish whether the
instructional method used in the treatment group (Instructional Method B) had a better effect on
students’ achievement in probability than the instructional method used in the control group
(Instructional Method A). Moreover, a comparison of post-test scores on the open-ended items
and a comparison of post-test total scores between the two groups were carried out.
In addition to an analysis of scores relative to achievement, analyses were performed to
determine the effects of the instructional treatment on students’ understanding of probability.
These analyses included i) a distractor analysis of multiple-choice items and ii) qualitative
analysis of transcripts of the conversations of the six students monitored in the control and
treatment groups. Recall that three students were selected from the control group and three from
the treatment group: one student of high mathematical ability, one student of moderate
mathematical ability and one student of low mathematical ability from each group. The selection
criterion was students’ mathematics grade in the last year of high school. The framework
developed by Jones, Thornton, Langrall and Tarr (1999) was used to qualitatively analyze the
transcripts. This framework included six constructs (sample space, experimental probability of
an event, theoretical probability of an event, probability comparisons, conditional probability,
196

and independence) across four levels of reasoning (Level 1: Subjective; Level 2: Transitional;
Level 3: Informal Quantitative; and Level 4: Numerical). Transcripts and their coding were
completed by the researcher and by a hired transcriber/second coder.
7.2 Summary of the Findings
7.2.1 Effects of Instructional Treatment on Students’ Achievement in Probability
Based on the results of the Wilcoxon Signed-Ranks test which was carried out to analyze
gain scores, the multiple-choice scores of students in the control group were significantly lower
on the post-test compared to the pre-test. The results indicated that for a majority of these items,
students in the control group had a lower performance on the post-test compared to the pre-test.
Moreover, all of the percentage descreases were higher compared to the percentage increases. In
particular, the most substantial descreases were with regards to Items 4, 11 and 13 in which
percent-correct responses decreased by 52.9%, 58.8% and 41.1% respectively .
In the case of the treatment group, student scores on the multiple-choice items did not differ
(increase) significantly from the pre-test to the post-test. Students in the treatment group had a
percent-correct increase on six multiple-choice items and a percent correct decrease on five
multiple-choice items. Moreover, most of the percentage increases were higher than the
percentage decreases. The most substantial increases were with regards to Items 3 and 10 in
which percent-correct responses increased by 25.9% and 48.2% respectively.
In addition to the analysis of raw gain scores for each of the two groups, a final piece of
analysis was carried out which involved normalized gain scores (Bao, 2006; Hakes, 1998) and
which directly addressed the research questions i.e. whether Instructional Method B (treatment)
had a better effect on students’ achievement on probability than Instructional Method A
(control). Positive as well as negative normalized gains existed in both groups. In particular, 12
197

students in the control group and 4 students in the treatment group exhibited negative normalized
gain scores. The Mann-Whitney test resulted in a p-value of 0.001 (< 0.05) indicating that the
normalized gain scores of the treatment group were significantly different from the normalized
gain scores of the control group. The test generated a mean rank of 14.41 for the control group
and a mean rank of 26.96 for the treatment group. This means that the control group had a bigger
number of lower normalized gain scores compared to the treatment group. Likewise, the
treatment group had a bigger number of higher normalized gain scores in comparison to the
control group. In summary, Instructional Method B had a significantly better effect on students’
achievement on probability than Instructional Method A.
Apart from the 14 multiple-choice items involved in the analysis of gain scores, the post-test
included an additional multiple-choice item on discrete probability distributions and two openended items: i) one on simple and joint probabilities, conditional probability and independence
and ii) one on the binomial distribution. With regards to this additional multiple-choice item,
23.5% of students in the control group and 48.1% of students in the treatment group responded
correctly. For the purposes of comparing the open-ended item scores of the control and treatment
groups, scoring rubrics were created which allotted numerical values to student responses.
The Mann-Whitney test was used to compare the results from the open-ended items in the
control and treatment groups. The test resulted in a p-value of 0.001 indicating that the scores of
the treatment group on the post-test open-ended items were significantly different from the
scores of the control group. The test generated a mean rank of 14.15 for the control group and a
mean rank of 27.76 for the treatment group. This means that the control group had a bigger
number of lower scores compared to the treatment group. Likewise, the treatment group had a
bigger number of higher scores in comparison to the control group. Therefore, the achievement
198

of students in the treatment group on the post-test open-ended probability items was significantly
higher than the achievement of students in the control group on the corresponding items.
Therefore, Instructional Method B was successful in producing significantly higher achievement
scores on the post-test open-ended items compared to Instructional Method A.
Non-parametric tests were also used to compare the total post-test scores of the control and
treatment groups. In particular, the Mann-Whitney test resulted in a p-value of 0.00 (< 0.05)
indicating that the post-test scores of the treatment group were significantly different from the
post-test scores of the control group. The test generated a mean rank of 12.53 for the control
group and a mean rank of 28.78 for the treatment group. This means that the treatment group had
a bigger number of higher scores compared to the control group. In summary, the post-test
achievement of students in the treatment group was significantly higher than the post-test
achievement of students in the control group. Therefore, Instructional Method B was successful
in producing significantly higher post-test scores compared to Instructional Method A.
7.2.2 Effects of Instructional Treatment on Students’ Understanding of
Probability
In this dissertation, students’ understanding of probability was measured through:
i)

a distractor (quantitative) analysis of student responses to the multiple-choice items on
the pre-test and post-test (Chapter 5) and

ii)

a qualitative analysis of audio-taped conversations as students worked in groups on
activities (treatment group) or problem sets (control group) (Chapter 6).
For the purposes of distractor analysis, item parts assessing the same heuristic or

misconception were grouped together. The percentages of students selecting the particular
distractor on each of the grouped items were added up and the mean percentage of students
199

applying the particular heuristic or misconception on the pre-test and post-test in each group was
computed.
In the case of the control group, the mean percentage of students who applied the outcome
approach remained stable at 5.9%; the same was true with regards to using the absolute size
instead of relative size when computing probabilities (percentage remained stable at a high
20.6%). In addition, there was a slight increase in the mean percentage of students who applied
the equiprobability bias which remained high on both the pre-test and the post-test (28.4% and
29.4% respectively). Moreover, there was a small increase in the mean percentage of students
who applied the negative recency (by 2.9%) and the positive recency (by 3.9%). The biggest
change in the case of the control group was in the application of the representativeness heuristic
(17.7% increase).
Unlike the results of the control group, the mean percentage of students in the treatment
group who applied the aforementioned heuristics or misconceptions mostly decreased. In
particular, the mean percentage of students who used the equiprobability bias, negative recency,
positive recency and outcome approach decreased by 3.4%, 9.25%, 4.9% and 1.85%
respectively. Only in the case of the representativeness heuristic and use of absolute size instead
of relative size when computing probabilities the mean percentage increased by 3.7% and 2%
respectively.
Considering students’ responses relative to other types of misconceptions represented by item
distractors, the mean percentage of students who believed that a larger number of outcomes in
the sample space implies a higher probability of occurrence increased in the case of the control
group (from 0% to 5.9%) and decreased in the case of the treatment group (by 3.7%). Second,
the mean percentage of students who applied division by an incorrect total when computing
200

probabilities using a two-way table increased by 3% in the case of the control group and by
11.1% in the case of the treatment group. Regarding the misconception that probability =
1/number of favorable outcomes when one item is selected at random, the mean percent of
students who applied this misconception remained stable at 5.6% in the case of the treatment
group but increased by 23.6% in the case of the control group.
Some misconceptions related to the union of events. The mean percentage of students who
believed that the term ‘or’ means considering only one of the events was higher in the case of the
control group on both the pre-test and post-test while in both groups the percentage of students
who applied this misconception slightly increased from pre-test to post-test. Similarly, the mean
percentage of students who applied the misconception that P( A  B) 

P( A)  P( B)
was higher
2

and remained stable at 29.4% in the case of the control group on both the pre-test and post-test
whereas it decreased by 7.4% in the case of the treatment group (from 14.8% to 7.4%). Unlike
the first two misconceptions relating to the union of events, the mean percentage of students who
applied the misconception P( A  B)  P( A)  P( B) decreased in the case of the control group
(by 5.9%) and increased in the case of the treatment group (by 3.7%).
For the purposes of qualitatively analyzing audio-taped data, transcripts of student
conversations of the six students monitored in the control and treatment groups as they worked in
groups were coded independently by the researcher and by a second rater. Once the coding of
transcripts was completed, they were checked by the researcher for agreement on the number of
excerpts relating to the constructs under study and the associated levels of reasoning. The
researcher and second coder met once more to discuss and resolve any instances of disagreement
in coding.
201

Recall that the framework used in this study does not include the concepts of discrete
probability distribution and binomial distribution. So, students’ level of reasoning as exhibited in
excerpts associated with discrete probability distributions (in Activity 3/Problem Set 3) and the
binomial distribution (in Activity 4 /Problem Set 4) could not be directly coded using the
framework. However, in Activity 3/Problem Set 3 and in Activity 4/Problem Set 4 students
needed to make connections to basic probability concepts; reasoning on such concepts is
discussed in the framework. Excerpts in which students made connections to basic probability
concepts were accounted for in the analysis of data relative to students’ understanding of basic
probability concepts. Any excerpts relative to the concepts of discrete probability distributions
and the binomial distribution in Activity 3/Problem Set 3 and in Activity 4/Problem Set 4
respectively in which students did not make connections to basic probability concepts were
briefly discussed in separate sections in chapter 6.
An examination of the results of the qualitative analysis regarding students’ reasoning levels
relative to sample space, theoretical probability, and experimental probability respectively,
revealed the following: i) The activities provided students with more direct as well as indirect
opportunities to reason about sample space and experimental probability compared to the
problem sets; and ii) The problem sets did not provide any direct or indirect opportunities for
students in the control group to reason about experimental probability and so, there was no data
on these students relative to this construct. Given these, it is not surprising that, as specified in
Chapter 5, students in the treatment group performed better on the multiple-choice items relating
to the Law of Large Numbers (Item 6 on the pre-test and post-test) and the concept of sample
space (Items 7 and 11 on the pre-test and post-test) than students in the control group. Specific to
Item 6 the percent-correct responses decreased by 11.7% in the control group and by 3.7% in the
202

treatment group from pre-test to post-test. On Item 7 the percent-correct responses increased by
5.8% in the case of the control group and by 18.6% in the treatment group from pre-test to posttest. Last, on item 11, the percent-correct responses decreased by a substantial 58.8% in the case
of the control group and increased by 3.7% in the treatment group. Considering these,
Instructional Method B (treatment) produced better results than Instructional Method A (control)
relative to experimental probability and the concept of sample space.
In addition, the results of the qualitative analysis regarding students’ reasoning levels relative
to sample space, theoretical probability, and experimental probability respectively, revealed that:
i) Students’ reasoning levels were the most stable with regards to sample space in comparison to
theoretical or experimental probability; ii) Students exhibited higher levels of reasoning with
regards to sample space (mostly L4) compared to theoretical or experimental probability across
the four activities and problem sets; and iii) with regards to experimental probability, the three
students monitored in the treatment group moved from a lower to a higher level reasoning in
Activities 1 and 4; moved from a higher to a lower level reasoning in Activity 2; and remained
stable at a low level reasoning in Activity 3.
With regards to theoretical probability, students in both groups mostly transitioned from a
lower to a higher level reasoning whereas in some cases students in the control or treatment
group exhibited a stable level of reasoning. In addition, students in the control group performed
0-3 transitions between levels of reasoning while students in the treatment group performed 0-2
such transitions. Based on these results, it is difficult to identify which instructional method had
a better effect on students’ understanding of theoretical probability.
An examination of the results of the qualitative analysis relative to the concepts of
conditional probability and independence revealed that both Activity 2 and Problem Set 2
203

provided students with direct opportunities to reason about conditional probability and
independence. Moreover, both of these instructional materials afforded opportunities in which
students needed to draw on previous knowledge of basic probability concepts (sample space,
theoretical probability of simple and joint events, experimental probability) in order to compute
conditional probabilities and decide whether events were independent. Overall, i) students’
reasoning with regards to conditional probability and independence was stable; ii) students
exhibited high levels of reasoning with regards to both concepts; iii) in the case of conditional
probability the low-ability and moderate-ability students in the treatment group demonstrated L4
reasoning whereas in the case of independence, all three students monitored in the control group
exhibited L4 reasoning; and iv) with regards to conditional probability all three students
monitored in the control group moved from L2 to L4 reasoning. Based on these results, it is
difficult to identify which instructional method had a better effect on students’ understanding of
conditional probability and independence.
In the case of discrete probability distributions, the three students monitored in the treatment
group were able to make connections to theoretical probability when computing the probability
distribution of a discrete random variable. Overall, the three students in the treatment group
exhibited L4 reasoning with regards to theoretical probability (and indirectly, with regards to
discrete probability distributions). In the control group, all three students were able to correctly
compute the expected value of a discrete random variable when a table listing the possible
outcomes and associated probabilities was provided. However, when asked to construct the
probability distribution themselves, all three students in the control group found such a task to be
challenging.

204

The last concept examined in this study was that of binomial distribution. The three students
monitored in the treatment group were able to construct the sample space for a random variable
that follows the binomial distribution thus exhibiting L4 reasoning with regards to sample space.
The high-ability and low-ability students in the treatment group demonstrated L4 reasoning with
regards to theoretical probability (and subsequently with regards to the binomial distribution)
since they were able to assign correct numerical probabilities to each outcome in the sample
space and then, correctly computed the probability of an event. The moderate-ability student in
the treatment group demonstrated L2 reasoning with respect to theoretical probability since the
student used quantitative judgments to compute the probability of each outcome but did so
incorrectly. In addition, the high-ability and moderate-ability students in the treatment group
were able to use a table of random numbers to calculate the experimental probability of an event
thus exhibiting L4 reasoning with regards to experimental probability (and subsequently with
regards to the binomial distribution). The treatment low-ability student was also able to use the
table of random numbers but instead of assigning a numerical probability the student provided a
frequency value for the experimental probability of the event thus demonstrating L3 reasoning
with respect to experimental probability (and subsequently with respect to the binomial
distribution). When it came to using the general formula for computing a probability from the
binomial distribution, a copy of each group’s written work indicated that students were able to
correctly compute the required probabilities from the binomial distribution. However, transcripts
of student conversations revealed that all three students monitored in the treatment group had
difficulties applying the general formula in the given context. The three students monitored in
the control group were able to define n (the number of observations in the sample space), X (the
number of events of interest), and p (the probability of an event of interest) as well as to compute
205

the mean and standard deviation of the binomial distribution. Moreover, all three students
monitored in the control group were able to apply the general formula for computing the
probability from the binomial distribution and assigned correct numerical probabilities as
responses to textbook problems.
7.3 Discussion of Findings
Based on the findings relative to students’ achievement in probability, scores on the multiplechoice items of students in the treatment group did not increase significantly from the pre-test to
the post-test. Surprisingly, the corresponding scores of students in the control group decreased
significantly from the pre-test to the post-test. Recall that the two groups were initially
equivalent with respect to pre-test multiple-choice item scores. Consideration of these results
leads one to wonder: Why were the post-test multiple-choice item scores of the control group
significantly lower compared to the pre-test multiple-choice item scores?
The results of the quantitative analysis provided in Chapter 5, revealed that in both the
control and treatment groups, student responses exhibited a ceiling effect on some items i.e.
more than 80% of the students in each group responded correctly to these items. In the treatment
group, student responses exhibited a ceiling effect on items 1, 2, 4, 11 and 12 on both the pre-test
and post-test, indicating that students in this group found the particular items to be easy. In the
case of the control group, a ceiling effect was observed on pre-test items 1, 4, 11, 13 and 14.
Interestingly, when it came to the post-test, there was a dramatic decrease in percent-correct
responses on items 4, 11 and 13 by 52.9%, 58.8% and 41.1% respectively. Overall, in the control
group, percent-correct responses decreased in nine out of the fourteen multiple-choice items that
were common to the pre-test and post-test. It should also be noted, that unlike the
aforementioned items, item 8 proved to be a very difficult item for students in both the control
206

and treatment groups since less than 30% of students in either group responded correctly to it on
the pre-test and post-test.
The information presented in the previous paragraph points towards the need for a closer
examination of the substantial decreases in percent-correct responses on items 4, 11 and 13 from
the pre-test to the post-test in the case of the control group. Considering item 4, in which students
were provided with a pictorial representation of all possible outcomes when two fair six-sided
dice are rolled and were asked to respond to the following,
You are about to roll 2 fair six-sided dice, hoping to get a double. (A double = both dice
show the same value on top). Which double will occur the least often?
a) 6 and 6
b) 1 and 1
c) 1 and 1, and, 6 and 6 are both least likely to occur.
d) All doubles are equally likely.
41.2% of students in the control group selected option c) as the correct answer on the post-test
whereas all students had correctly selected option d) on the pre-test. A possible explanation for
this phenomenon might lie in the instructional materials used in the two groups in this study. The
activities used in the treatment group involved experiments in which students rolled two or three
fair dice and computed probabilities of various events. However, the problem sets provided in
the course textbook on which students in the control group worked in groups during class did not
provide any hands-on opportunities for students to work with dice and so, these students were
not provided with as many opportunities to overcome their misconceptions relating to dice
outcomes. Students’ selection of option c) is supported by results of other studies reported in the
literature (Green, 1983; Konold et al., 1993) in which participants tended to prefer middle
numbers on a die and avoid the numbers 1 and 6, thinking that the middle numbers have a higher
chance of occurring.
207

The next item under consideration is item 11 in which students were asked to respond to
the following:
The eleven chips shown below are placed in a bag and mixed. Chelsea draws
one chip from the bag without looking.
Figure 7.1
Pre/Post-Test Item 11 Coin Figure

What is the probability that Chelsea draws a chip with a number that is a
multiple of three?
1
a)
11
1
b)
3
4
c)
11
4
d)
7
On the pres-test, 94.1% of students in the control group had correctly selected option c).
Surprisingly, on the post-test, 47.1% of these students selected option a) (whereas none of them
had made this selection on the pre-test). A possible explanation may be that problems in the
course textbook tend to reinforce the idea of equally likely outcomes and so, as reported in the
literature, students believed that in order for an experiment to be “fair”, all outcomes needed to
be equally likely (Jacobs, 1999; Shaughnessy, 2003). Another explanation might involve the
issue of language. When responding to this item on the pre-test during regular class time, many
participants (especially in the control group) were not aware of the meaning of ‘multiple of’ and
asked for a translation of the term in Greek at which time I (instructor/researcher) wrote the
208

translation on the classboard. Recall that the post-test was embedded in the course final exam
which did not take place during regular class time. Students had the right to ask a supervisor
during the final exam for a translation of a few words in Greek. It is possible that many students
in the control group could not remember the meaning of ‘multiple of’, did not ask an invigilator
for its meaning, and instead focused on the idea of drawing one out of the eleven chips in the
bag.
The third item in which percent-correct responses exhibited a substantial decrease from
pre-test to post-test was item 13 which is presented below:
The figure below shows a spinner with 24 sectors. When someone spins the
arrow, it is equally likely to stop on any sector.
Figure 7.2
Pre/Post-Test Item 13 - Spinner

3 of the sectors are blue, 1 is purple, 12 are orange, and 8 are red.
If a person spins the arrow, on which color sector is the spinner LEAST likely to stop?
a)
Blue
b)
Purple
c)
Orange
d)
Red
Relative to this item, 35.3% of students in the control group selected option c) on the post-test. It
seems that students in the control group confused the meaning of ‘least’ with the meaning of
‘most’ likely when responding to this item, pointing towards students’ difficulties with the
English language.
209

Recall that participants in the introductory statistics course in which this study took place
were split into three sections by the course coordinator in consultation with the academic board
at the college, based on students’ English language proficiency. Two of the sections served as the
treatment group and one section as the control group. One of the treatment sections included
students of moderate/high English language proficiency whereas the other included students of
all levels of English language proficiency. The control section included students of low English
language proficiency. Course instructors were not aware of this type of placement until the
commencement of the academic year. Moreover, due to this placement, it is challenging to
identify the extent to which the results are attributed to difficulties with the English language or
the instructional method. As noted in items 11 and 13, possible explanations for substantial the
decreases in percent-correct responses in the case of the control group may be (at least partially)
explained by difficulties with the English language. Researchers suggest that linguistic
difficulties associated with the terminology used for probability are prevalent among students
(Green 1982a, Konold, 1988; see Ulep, 1990). Adding to this, students’ English language
difficulties, creates further obstacles to students’ understanding and achievement in probability.
Further research needs to be carried out in order to examine the factors that might have led to
these changes in multiple-choice item scores. The fact that students completed the pre-test during
a regular classroom session whereas the post-test was administered under exam conditions (it
was embedded in the course final exam) in combination with the low English proficiency of
students in the control group, might have led to the significant decrease in post-test multiplechoice item scores in the case of the control group. Exam conditions may add to students’ stress
level and in combination with English language difficulties may impede their comprehension of
probability items stated in English. The moderate/high English language proficiency of students
210

in one of the treatment sections may have helped students better demonstrate their actual
understanding of probability concepts since such level of proficiency might have helped students
better comprehend the context of each multiple-choice item. In this manner, students having less
difficulties with the English language might make the results of the two treatment sections more
directly attributable to the instructional treatment used.
Similarly, in the case of the open-ended items included on the post-test, the scores of the
treatment group were significantly higher than the scores of the control group. As in the case of
the multiple-choice items, here as well, the instructional method used and the language issue
might have both played a role in the results exhibited by participants. The activities used in the
treatment group included both a hands-on experiment and a theoretical component so students
were given the opportunity to study a concept in both manners and make connections between
the experimental and theoretical approach to probability. Such an approach has been shown to
aid students in overcoming probability misconceptions (Shaughnessy, 1977, 1981) and has been
supported by researchers (Steinbring, 1984; 1991; see Jones and Thornton, 2005). The problem
sets that students in the control group worked on did not include an experimental component; in
these problems students needed to apply knowledge they had gained during lectures on a
probability concept to merely solve the problem at hand. However, according to Shaughnessy
(1981), “an initial formalistic approach to probability is unlikely to help students overcome
misconceptions” (p. 95). The fact that the course textbook did not provide students in the control
group with direct opportunities to work with experimental probability indicated a preference for
a classical approach to probability. This comes in contrast to current calls to teach probability in
conjunction with data collection and analysis (Shaughnessy, 2003). Moreover, the low English
language proficiency of students in the control group might have led to difficulties in
211

comprehending the problem context and in turn might have caused the significantly lower openended item results in comparison to the results of the treatment group on the corresponding
items.
Results of the distractor analysis relative to students’ understanding of probability are in line
with research on students’ probabilistic misconceptions. As evident in the literature, probabilistic
misconceptions are resistant to change (Konold, 1995) and in general, “there is no simple story
about how students reason about chance” (Konold et al., 1993, p. 413). Specific to the distractor
analysis carried out in this study, the results indicated that the percentage of students applying
some of the probabilistic heuristics, increased in some cases whereas it decreased in others. This
was true in both the treatment and control groups. However, with regards to the 12 heuristics
presented in Tables 5.9 and 5.10 in Chapter 5, the percentage of students in the control group
who applied these increased in the case of 8 (67%) of them, remained stable in the case of 3
(25%) of them and decreased in the case of 1 (8%) of them. On the other hand, the percentage of
students in the treatment group who applied these heuristics, increased in the case of 5 (42%) of
them, remained stable in the case of 1 (8%) of them, and decreased in the case of 6 (50%) of
them. These results are in line with past research that indicated that the use of activity-based
instruction may help students with respect to probabilistic misconceptions (Shaughnessy, 1977,
1981).
7.4 Strengths and Limitations
7.4.1 Strengths
Although the sample of students who participated in this study cannot be regarded as
representative of statistics students in general, the reasoning and understanding abilities of these
students and the role that different types of tasks have on their learning and understanding of
212

probability may aid educators gain knowledge about students who study statistics in a collegelevel service course.
The fact that the class sections (treatment and control) were taught at the same setting, by
the same instructor, using the same textbook and language of instruction, may be considered
strengths. Controlling for these factors allowed the results of the study to be more clearly
attributed to the different tasks used in the study, thus more directly addressing the research
questions.
For the purposes of data collection, audio recorders were used to record verbatim the
responses students gave as they worked in groups on probability problems or activities which
involved open-ended probes or questions. Using audio recordings as a research tool allows the
researcher to obtain more data and to have easy reach to students’ verbal interactions and
responses thus aiding in addressing the research questions. Given that the researcher was also the
instructor it would have been impossible to keep track of students’ group discussions and
reasoning as they worked in class without a means of data collection such as audio recording.
Related to the above issue, the fact that the instruments included both multiple-choice and
open-ended questions can be considered an advantage of the study. Each of these two types of
questions has advantages and disadvantages. Although multiple-choice questions do not allow
respondents to express their reasoning freely, open-ended questions do so. Moreover, it is more
challenging to analyze open-ended questions however, multiple-choice items are easier and
quicker to quantify. Therefore, using a combination of multiple-choice and open-ended items on
the instruments adds to the strength of the study.

213

7.4.2 Limitations
A limitation of the study is that the interpretation of the data might involve a degree of
bias since the instructor was also the researcher (Zimmerman, 2002). In order to control for such
bias a second coder was hired; this coder helped code participant responses to questions on the
activities or in the problem sets.
Second, “[A] criticism of qualitative research methods is that it is very difficult to make
and justify generalizations that apply to other settings.” (Jaworski, 1998, p. 119). This study was
performed in a single college at a particular town in Cyprus. More diverse student populations
may respond differently than the participants in this study. Therefore, the results may only be
generalized to students with similar demographics.
Third, the design of the study was complicated by the fact that students were placed in
the course sections by the course administrator and academic board based on their English
language proficiency. This placement caused obstacles in the interpretation of results since it was
challenging to determine the extent to which the study’s results were attributable to the method
of instruction or to language issues.
Last, the search for a framework to be used in analyzing students’ understanding of
probability proved to be a very challenging task. Any available frameworks I came across to
related to grades K-12 (Jones et al., 1997, 1999; Jones and Thornton, 2005). In this study, the
framework used to analyze students’ understanding of probability was that by Jones, Thornton,
Langrall and Tarr (1999) which included six constructs (sample space, experimental probability
of an event, theoretical probability of an event, probability comparisons, conditional probability,
and independence) across four levels of reasoning (Level 1: Subjective; Level 2: Transitional;
Level 3: Informal Quantitative; and Level 4: Numerical). The framework is limited in that i) it
214

did not directly relate to college students’ understanding of probability; and ii) it did not
accommodate for all probability constructs under examination i.e. discrete probability
distributions and the binomial distribution. Moreover, the type of classification used in the
framework assumes that i) “an individual's thinking about a specific probabilistic construct will
be consistent across different contexts related to that construct” and ii) “that an individual's
thinking will be consistent across the constructs themselves” (Rubel, 2007). However, as
evident from the results, students’ reasoning regarding the probability constructs under
examination did not model such consistencies. This lack of consistency in reasoning is evident in
the tables presented in Chapter 6 which reveal that students transitioned between levels of
reasoning during the activities or problem sets. Similar to this study, Jones et al. (1999) also
came across “these types of instability within their data, especially following instruction, and
speculated that students' probabilistic thinking might be tightly woven within the features of
the context itself” (Rubel, 2007). Although the framework by Jones et al. (1999) has its
limitations, it was selected because it distinctly addressed each of the constructs of experimental
and theoretical probability in a cognitive manner (thus aiding in addressing the issue of
understanding) and these two constructs were vital elements of the study.
7.5 Implications and Recommendations for Further Research
The findings of this study may help mathematics instructors and curriculum developers
by providing valuable information on the effect of particular tasks on students’ achievement and
understanding of probability. While the sample used in this study may not be viewed as
representative of statistics students in general, the perceptions of these participants may help
educators understand tertiary level students who study probability as part of a service course
rather than by choice.
215

As the Curriculum Principle (NCTM, 2000) states, curriculum materials should
emphasize the connections between mathematical ideas. An examination of the course textbook
used in the course in which this study took place, indicated that it did not afford students with
opportunities to study experimental probability, particularly through modeling. However,
researchers (e.g. Steinbring, 1984; 1991; see Jones and Thornton, 2005; Shaughnessy, 2003)
recommend that instruction in probability should help students view the connections between
theoretical and experimental probability. Therefore, curricular materials should involve the
modeling of real-world situations through simulations. Moreover, during the process of selecting
the instructional materials to be used in this study, I realized that there is a lack of available
activities that could be used during instruction on probability in introductory statistics courses at
the tertiary level. Given this, curriculum developers should ensure that such activities are
developed.
In Cyprus, research on students’ achievement and understanding of probability has not
been emphasized, especially at the secondary and post-secondary levels. Furthermore, the role of
such research on the teaching and learning of probability, and in curriculum development has not
been widely applied. Current practices are under review as reform initiatives are being
implemented in mathematics classrooms at all levels (see Chapter 2). Reports on the reformed
mathematics curriculum place increased emphasis on new topics that have rarely been taught at
the school level in Cyprus, including probability, and in the way these topics are taught i.e. using
increased student involvement in the learning process, cooperative learning and technology
(Papastylianou, 1997; see also Papanastasiou, 2002; Vrasidas and McIsaac, 2001). In making
such changes to the curriculum, it is important to be aware of available research. The findings of
this study suggest that curriculum developers and teachers promote the use of activity-based
216

instruction that helps increase students’ achievement on probability. Teachers and curriculum
developers should make use of real data and activities as means in encouraging students to
express their reasoning and modify their beliefs with regards to probability concepts (Garfield &
Ahlgren, 1988; Shaughnessy, 1992).
The results of this study also inform my future probability instruction. If another
opportunity arose to teach a college-level course in probability in which I would have total
control of student placement in the various course sections (unlike the situation that arose in this
study), then, based on these results I would i) assign students to course sections randomly and
not based on their English language proficiency; and ii) use the instructional method
implemented in the treatment group in this study since it had a positive effect on students’
achievement in probability.
The results of this study give rise to recommendations for further research. First, let us
consider research that could be carried out as a follow-up to this study. The findings relative to
students’ achievement indicate that the instructional treatment of using activities that generate
real data while having students work in small groups during probability instruction, significantly
increased students’ achievement in probability. Further research needs to be carried out with the
aim of revealing the factors that help explain these significant differences in achievement
between the treatment and control groups in this study. Factors such as gender, the role of
language, students’ confidence in mathematics, and students’ feelings regarding the use of group
work should be examined. The limitations identified relative to the framework used in this
dissertation give rise to the following needs: i) the development and validation of a framework
that can be used to assess college level students’ understanding of probability (including the
construction of a clear definition of the term probabilistic understanding); ii) the development
217

and validation of a framework that can be used to assess students’ understanding of discrete
probability distributions and the binomial distribution. In addition, findings from a larger group
of students at various tertiary education settings could produce trends that were not evident in
this study. In addition, testing a larger sample of students could provide more insight into the
prevalence of the resulting trends.
Particular to the language factor and given that instruction at the research site was carried
out in English whereas the students’ first language was Greek, further research is needed into the
relationship between classroom discourse relative to probability carried out in Greek and that in
English, and whether students’ difficulties arise from these relationships. Relative to this issue,
the effectiveness of allowing students to discuss probability in their home language during
instruction, needs to be investigated as well as the role played by language in the assessment of
achievement in probability; in particular, the effects of tests written in the students’ first
language on achievement in probability should be investigated.
A question that remains is the role that additional reflection might play in students’
understanding of probability from the activities. In this study students were not directly assigned
any out of class work related to the activities nor were they directly assessed on exams on the
content presented in the activities using questions that resembled the activities. So, one may
assume that students spent a minimal amount of time outside of class reflecting on the activities.
The question remains as to whether additional reflection might help students develop and retain
an understanding of the probability concepts presented in the activities.

218

APPENDICES

222

APPENDIX A: CONSENT FORM AND PERMISSION FOR USE OF REPRINTS

MICHIGAN STATE
U N I V E R S I T Y
1/15/10
Title of Research Study: College Students’ Achievement and
Understanding of Experimental and Theoretical Probability: The Role of
Tasks

Dear Student:
I am contacting you to ask whether you would be willing to participate in a
research project relating to mathematics education. In particular, the
project attempts to examine the role of different tasks on college students’
achievement and understanding of probability.
If you agree to participate, then I would kindly ask that you fill out a
background questionnaire, a pre-test and post-test of probability problems.
You may also be asked to work on a set of classroom group activities or
problem sets. As you will be working on the group activities your group
conversations will be audio-taped and these tapes will be transcribed by me
after the final course grades have been completed and submitted.
Division of Science and
Mathematics Education
Michigan State University
North Kedzie Hall
East Lansing, MI
48824-1034

Such a study holds benefits as well as some risks for participants.
Educational research has revealed that active learning and communication
in mathematics classrooms can strengthen students' reasoning and clarify
their understanding of mathematics concepts. Moreover, research has
shown that students engage more with the material and more of them are
able to complete the introductory statistics course when cooperative
learning is used during instruction. Such instructional methods will be part
of this study and so, you may benefit through your participation. However,
there is also some small risk in that you might feel embarrassed if you do
not know the answer to a problem or do not know how to proceed with
some step of a probability experiment which you will be performing as part
of this study. Also, thinking out loud and/or being in a cooperative learning
environment in the classroom might be an unfamiliar process to you and
you might feel somewhat uncomfortable at first. Always feel free to
express any difficulties you encounter as your work on the group activities
and to seek help from your classmates or your instructor.
223

The results of this study might be published in scientific research journals
and professional publications for teachers. Only I will use the tapes or
transcripts for analysis and that will be done after final grades have been
completed and submitted. Access to the tapes and transcripts will be
protected according to university regulations and up to the maximum
extent allowable by law. No data that is collected will identify the student
from which it was collected, other than to members of this research team.
Results of the study will be made available to you if you request them.
Participation in this study is voluntary and you may withdraw from
participation at any time. Your consent to participate or your decision to
withdraw will not be made known to me until after the final course grades
have been completed and submitted.
The data materials (background questionnaire, pre-test and post-test) will
be collected by the course coordinator. I will not examine these materials
until after the final course grades have been completed and submitted. If
you decide not to participate or if you withdraw from the study at any
point, the data materials you complete as well as anything you say that
might have been recorded on tape will not be used as data for this study.
If you are willing to participate, please sign this form and return it to the
course coordinator.
Student’s Printed Name ________________________________________
Student’s Signature ____________________________________________
Date____________

If you have any questions about this project, please do not hesitate to
contact the researcher, Irini Papaieronymou, by phone (99784507), e-mail
(papaiero@msu.edu) or regular mail: Corner of Faneromenis Ave. and
Kalvou Str., P.O.Box 40763, 6307 Larnaca, Cyprus. If you have any
questions or concerns about your rights as a participant in this study, you
may contact Michigan State University's University Committee on
Research Involving Human Subjects (UCRIHS) at 202 Olds Hall,
Michigan State University, East Lansing, MI 48824-1046, Phone: (517)
355-2180, E-Mail:irb@msu.edu.
Sincerely,
Irini Papaieronymou
PhD Candidate in Mathematics Education, Michigan State University,USA
Mathematics Instructor, PA College, Larnaca, Cyprus
224

AUDIOTAPE CONSENT STATEMENT
Cooperative Learning in a Probability Project
One of the goals of this project is to help teachers and researchers better understand students'
probabilistic thinking. To do this, we must give examples of students' work. One way to give
examples is to provide transcripts of what students say as they are solving problems (with
students not being identified, of course). If you would be willing for your audio-taped work to
be shared with teachers and researchers, please give your consent below.
I agree that portions of audiotapes of my conversations with members of my group as we work
on probability problems or experiments during class time may be used by Irini Papaieronymou in
presentations and classes for teachers or researchers. I understand that I have the right to listen to
the audiotapes before they are used. I have decided that I:
____want to listen to the audiotapes
____do not want to listen to the audiotapes
Sign now if you do not want to listen to the audiotapes. If you want to listen to the tapes, you
will be asked to sign after listening to them.

___________________________
Student’s Printed Name

_

___________________________
Student's Signature

_

_________________________
Date

225

PERMISSION FOR USE OF REPRINTS (ARTIST PROJECT)
Dear Irini:
Thank you for registering for the Web ARTIST Assessment Builder. We hope you have had a
chance to review and download some of our items.
We offer 11 online multiple choice tests on individual topics, and the CAOS (Comprehensive
Assessment of Outcomes in a first Statistics Course) test. We encourage you to try one or more
of our online topic tests which have been found to be especially helpful for reviewing material.
We also encourage you to use the CAOS as a pretest and posttest to measure gains in student
learning. All test results can be quickly and easily downloaded once students have completed
taking each test. For more information and to sign up to administer a test, please go
to: https://app.gen.umn.edu/artist/tests/index.html
We invite you to send us copies of any good assessment items or materials that you are willing
to contribute to our website by sending them in an e-mail to Bob delMas (delma001@umn.edu).
We have a Question and Answer page where we post questions we receive that we think will be
of interest to users, and we share our responses to these questions. So feel free to send
questions.
Thanks in advance for your feedback, and let us know if you have any problems or questions.

Best regards,
Joan Garfield, Bob delMas, Beth Chance, and Ann Ooms
The ARTIST Team
---------------------------------------------------------------------------------------------------------------------

226

Irini:
Now that you are registered to access the online tests, you can have a PDF file of the CAOS test
sent to you. You can make copies of this to use in class. We ask that you collect the tests from
the students and do not let students keep copies of the CAOS test. There is no need to send us an
Excel file of the students' responses.
You can learn more about retrieving a PDF file of the CAOS test at:
https://app.gen.umn.edu/artist/tests/index.html
Read the instructions for Step 2.

Best regards,
Bob delMas
*******************************
Robert C. delMas, Ph.D.
Associate Professor
Quantitative Methods in Education
Department of Educational Psychology
University of Minnesota
Phone: (612) 625-2076
Fax: (612) 624-8241

227

Irini:
Yes, you can definitely use the ARTIST Probability Scale for your dissertation research. For
your Reference section, you can list the project as follows.
Garfield, J., delMas, R., Chance, B. (2002). Assessment Resources Tools for Improving
Statistical Thinking (ARTIST). A National Science Foundation funded project (CCLI-ASA0206571). https://app.gen.umn.edu/artist/
And you can cite it as Garfield, delMas and Chance (2002) in the body of your dissertation.
Bob

228

APPENDIX B: INSTRUMENTS

B.1 BACKGROUND QUESTIONNAIRE

Student Background Questionnaire

A. Background Information:

1. First Name ______________________

Last Name ______________________

2. Age in years ________

3. Town/Country of birth ____________________

4. What is your first language? (Check one)
___ Greek
___ English
___ Other (please specify) ____________

229

5. Please indicate your level of English proficiency (Circle one in each column):
Table B.1
English language proficiency level
Language
English

Oral
Poor Good

Written
Excellent Poor Good

Reading

Excellent Poor Good

Excellent

B. Educational Background:

6. Which of the following best describes the high school you graduated from?
_____ Public high school in Cyprus (continue to question 7)
_____ Private high school in Cyprus (continue to question 8)
_____ I did not graduate from a high school in Cyprus. I graduated from a high school in
(specify country) ______________

7. If you attended a public high school in Cyprus did you take advanced mathematics?
___ Yes
___ No

8. What was your grade in mathematics in your last year of high school? ________

230

9. Did you ever study statistics and probability in high school?
____ Yes
____ No

10. What year are you at P.A. College? (Check one)
st

___Freshman (1 year student)
___Sophomore (2

nd

year student)

rd

___Junior (3 year student)
th

___Senior (4 year student)

11. What is your major at P.A. College? (Check one)
___ Accounting (4 years, Bachelor of Arts)
___ Business Administration (4/5 years, Bachelor of Arts)
___ Business Computing (4/5 years, Bachelor of Science)
___ Banking (4 years, Bachelor of Arts)
___ Business Administration (2 years, Diploma)
___ Computing and Information Systems (2 years, Diploma)
___ Business and Information Technology (1 year, Certificate)
___ Law (1 year, Certificate)

231

B.2 PRE-TEST
Probability Pre-Conceptions
Please work individually on the following problems. In each problem, circle the answer that
seems most reasonable to you. Your answers (whether correct or incorrect) will help your
instructor to determine what you already know about probability, prior to instruction in this
course on probability.

1. George and Mike each bought one ticket for a lottery each week for the past 100 weeks.
George has not won a single prize yet. Mike won 20 euros last week, for the first time. Who is
more likely to win a prize this coming week if they each buy only one ticket?
a)

George

b)

Mike

c)

They have an equal chance of winning

2. Suppose you read on the back of a lottery ticket that the chances of winning a prize are 1 out
of 10. Select the best interpretation.
a)

You will win at least once out of the next 10 times you buy a ticket.

b)

You will win exactly once out of the next 10 times you buy a ticket.

c)

You might win once out of the next 10 times but this will not happen for sure.

232

3. Melita is flipping a fair coin with her eyes shut. Nora records the outcome and places the coin
again in Melita’s hand. Heads has just come up 5 times in a row! The chance of getting heads
on the next flip is
a)

less than the chance of getting tails since we are expected to get tails.

b)

equal to the chance of getting tails since the flips are independent and the coin is fair.

c)

greater than the chance of getting tails since heads seem to be coming up.

For problems 4 and 5 consider the following diagram which shows the 36 possible results when
two six-sided dice are rolled.
Figure B.1
Pre-Test Items 4 and 5 – Sample Space for Rolling Two Fair Dice

4. You are about to roll 2 fair six-sided dice, hoping to get a double. (A double = both dice
show the same value on top). Which double will occur the least often?
a) 6 and 6
b) 1 and 1
c) 1 and 1, and, 6 and 6 are both least likely to occur.
d) All doubles are equally likely.
233

5. Maria rolls two six-sided dice at the same time. Each side of each die is uniquely labeled with
a number from 1 to 6. The following are two of the possible results that could occur when
these two dice are rolled:

Result 1: a 5 and a 6 are obtained in any order.
Result 2: a 5 is obtained on each die.
Which of the following statements is correct?
a)

The probability of obtaining each of these results is equal.

b)

There is a higher probability of obtaining Result 1 (a 5 and a 6 in any order).

c)

There is a higher probability of obtaining result 2 (a 5 on each die).

d)

It is impossible to give an answer.

6. A game company created a little plastic dog that can be tossed in the air. It can land either
with all four feet on the ground, lying on its back, lying on its right side, or lying on its left
side. However, the company does not know the probability of each of these outcomes. They
want to estimate the probabilities. Which of the following methods is most appropriate?
a) Since there are four possible outcomes, assign a probability of ¼ to each outcome.
b) Toss the plastic dog many times and see what percent of the time each outcome occurs.
c) Simulate the data using a model that has four equally likely outcomes.

234

7. A set of 24 cards is numbered with the positive integers from 1 to 24. The cards are shuffled
and only one card is selected at random. What is the probability that the number on the card
can be divided by 4 or 6?
a)

1
6

b)

5
24

c)

1
4

d)

1
3

e)

5
12

235

8. Two containers, labeled A and B, are filled with red and blue marbles according to the
quantities listed in the table below. Each container is shaken several times and marbles are
then selected from the containers at random, without looking.
Table B.2
Pre-Test Item 8 – Marbles in Container
Container

A

B

Red

80

40

Blue

20

60

Which of the following outcomes has the smallest probability?
a) Obtaining a blue marble from container A.
b) Obtaining a blue marble from container A and a blue marble from B.
c) All of the above are equally likely.

236

9. Two containers, labeled A and B, are filled with red and blue marbles according to the
quantities listed in the table below. Each container is shaken vigorously. After choosing one
of the containers, you will reach in and, without looking, draw out a marble. If the marble is
blue, you win 50 euros.
Table B.3
Pre-Test Item 9 – Marbles in Container
Container

A

B

Red

6

60

Blue

4

40

Which container gives you the best chance of drawing a blue marble?
a) Container A (with 6 red and 4 blue)
b) Container B (with 60 red and 40 blue)
c) Equal chances from each container

237

10. One thousand people selected at random were questioned about smoking and drinking. The
results of this survey are summarized in the table below.
Table B.4
Pre-Test Item 10 Cross Tabulation
Smokers

Non-smokers

Drinkers

320

530

Non-drinkers

20

130

What is the probability that a randomly selected respondent drinks and smokes?
a)

320
340

b)

320
850

c)

320
1000

d)

1
320

238

11. The eleven chips shown below are placed in a bag and mixed. Chelsea draws one chip from
the bag without looking.
Figure B.2
Pre-Test Item 11 Coin Figure

What is the probability that Chelsea draws a chip with a number that is a
multiple of three?
a)

1
11

b)

1
3

c)

4
11

d)

4
7

239

12. Sophie has a bag in which there are 16 marbles: 8 are red and 8 are black marbles. She
shakes the bag and without looking she draws 2 marbles from it and does not put them back.
Both marbles that she has drawn are black. She then draws a third marble out of the bag.
What can you say about the likely color of this third marble?
a) It is more likely to be red than black.
b) It is more likely to be black than red.
c) It is equally likely to be red or black.
d) You cannot tell if red or black is more likely.

13. The figure below shows a spinner with 24 sectors. When someone spins the arrow, it is
equally likely to stop on any sector.
Figure B.3
Pre-Test Item 13 -Spinner

3 of the sectors are blue, 1 is purple, 12 are orange, and 8 are red.
If a person spins the arrow, on which color sector is the spinner LEAST likely to stop?
a)

Blue

b)

Purple

c)

Orange

d)

Red
240

14. The smaller box contains 20 tickets numbered from 1 to 20. The larger box contains 100
tickets numbered from 1 to 100.
Figure B. 4
Pre-Test Item 14 – Ticket Boxes

20 tickets

100 tickets

Without looking at them you can pick a ticket from either box. Which box would give
you the greater chance of picking out a ticket with the number 17 on it?
a)

The box with 20 tickets.

b)

The box with 100 tickets.

c)

Both boxes would give the same chance.

d)

It is impossible to tell.

241

Sources:
Garfield, J., delMas, R., Chance, B. (2002). Assessment Resources Tools for Improving
Statistical Thinking (ARTIST). A National Science Foundation funded project (CCLI-ASA0206571).
Mullis, I.V.S., Martin, M. O., Beaton, A. E., Gonzalez, E. J., Kelly, D. L., and Smith, T. A.
(1998). Mathematics and Science Achievement in the Final Year of Secondary School:
IEA’s Third International Mathematics and Science Study (TIMSS). Chestnut Hill, MA:
Boston College.
International Association for the Evaluation of Educational Achievement. TIMSS 1999: IEA’s
repeat of the Third International Mathematics and Science Study at the Eighth Grade:
TIMSS mathematics items. Chestnut Hill, MA: Boston College.
International Association for the Evaluation of Educational Achievement (2009). TIMSS 2007
user guide for the international database. TIMSS and PIRLS International Study Center,
Lynch School of Education, Boston College.
International Association for the Evaluation of Educational Achievement (2007). TIMSS 2003:
Mathematics items released set: Eighth grade. TIMSS and PIRLS International Study
Center, Lynch School of Education, Boston College.

242

B.3

POST-TEST

1.

George and Mike each bought one ticket for a lottery each week for the past 100 weeks.
George has not won a single prize yet. Mike won 20 euros last week, for the first time.
Who is more likely to win a prize this coming week if they each buy only one ticket?
a)
b)

Mike

c)

2.

George

They have an equal chance of winning

Suppose you read on the back of a lottery ticket that the chances of winning a
prize are 1 out of 10. Select the best interpretation.
a)
b)

You will win exactly once out of the next 10 times you buy a ticket.

c)

3.

You will win at least once out of the next 10 times you buy a ticket.

You might win once out of the next 10 times but this will not happen for sure.

Melita is flipping a fair coin with her eyes shut. Nora records the outcome and
places the coin again in Melita’s hand. Heads has just come up 5 times in a
row! The chance of getting heads on the next flip is
a)

less than the chance of getting tails since we are expected to get tails.

b)

equal to the chance of getting tails since the flips are independent and the coin is
fair.

c)

greater than the chance of getting tails since heads seem to be coming up.

243

For problems 4 and 5 consider the following diagram which shows the 36 possible results
when two six-sided dice are rolled.
Figure B. 5
Post-Test Items 4 and 5 – Sample Space for Rolling Two Fair Dice

4. You are about to roll 2 fair six-sided dice, hoping to get a double. (A double = both dice
show the same value on top). Which double will occur the least often?
a) 6 and 6
b) 1 and 1
c) 1 and 1, and, 6 and 6 are both least likely to occur.
d) All doubles are equally likely.

244

5. Maria rolls two six-sided dice at the same time. Each side of each die is uniquely labeled
with a number from 1 to 6. The following are two of the possible results that could occur
when these two dice are rolled:
Result 1: a 5 and a 6 are obtained in any order.
Result 2: a 5 is obtained on each die.
Which of the following statements is correct?
a) The probability of obtaining each of these results is equal.
b) There is a higher probability of obtaining Result 1 (a 5 and a 6 in any order).
c) There is a higher probability of obtaining result 2 (a 5 on each die).
d) It is impossible to give an answer.

6.

A game company created a little plastic dog that can be tossed in the air. It can land either
with all four feet on the ground, lying on its back, lying on its right side, or lying on its
left side. However, the company does not know the probability of each of these
outcomes. They want to estimate the probabilities. Which of the following methods is
most appropriate?
a) Since there are four possible outcomes, assign a probability of ¼ to each outcome.
b) Toss the plastic dog many times and see what percent of the time each outcome
occurs.
c) Simulate the data using a model that has four equally likely outcomes.

245

7. A set of 24 cards is numbered with the positive integers from 1 to 24. The cards are
shuffled and only one card is selected at random.
What is the probability that the number on the card can be divided by 4 or 6?
a)

1
6

b)

5
24

c)

1
4

d)

1
3

e)

5
12

246

8. Two containers, labeled A and B, are filled with red and blue marbles according to the
quantities listed in the table below. Each container is shaken several times and marbles
are then selected from the containers at random, without looking.
Table B.5
Post-Test Item 8 – Marbles in Container
Container

A

B

Red

80

40

Blue

20

60

Which of the following outcomes has the smallest probability?
a)

Obtaining a blue marble from container A.

b)

Obtaining a blue marble from container A and a blue marble from B.

c)

All of the above are equally likely.

247

9. Two containers, labeled A and B, are filled with red and blue marbles according to the
quantities listed in the table below. Each container is shaken vigorously. After choosing
one of the containers, you will reach in and, without looking, draw out a marble. If the
marble is blue, you win 50 euros.
Table B.6
Post-Test Item 9 – Marbles in Container
Container

A

B

Red

6

60

Blue

4

40

Which container gives you the best chance of drawing a blue marble?
a) Container A (with 6 red and 4 blue)
b) Container B (with 60 red and 40 blue)
c) Equal chances from each container

248

10.

One thousand people selected at random were questioned about smoking and
drinking. The results of this survey are summarized in the table below.
Table B.7
Post-Test Item 10 Cross Tabulation
Smokers
Drinkers

130

320
1000

d)

20

320
850

c)

530

320
340

b)

320

Non-drinkers

a)

Non-smokers

1
320

249

11.

The eleven chips shown below are placed in a bag and mixed. Chelsea draws
one chip from the bag without looking.
Figure B.6
Post-Test Item 11 Coin Figure

What is the probability that Chelsea draws a chip with a number that is a
multiple of three?
a)

1
11

b)

1
3

c)

4
11

d)

4
7

250

12.

Sophie has a bag in which there are 16 marbles: 8 are red and 8 are black marbles. She
shakes the bag and without looking she draws 2 marbles from it and does not put them back.
Both marbles that she has drawn are black. She then draws a third marble out of the bag.
What can you say about the likely color of this third marble?
a) It is more likely to be red than black.
b) It is more likely to be black than red.
c) It is equally likely to be red or black.
d) You cannot tell if red or black is more likely.

13.

The figure below shows a spinner with 24 sectors. When someone spins the arrow, it is
equally likely to stop on any sector.
Figure B.7
Post-Test Item 13 - Spinner

3 of the sectors are blue, 1 is purple, 12 are orange, and 8 are red.
If a person spins the arrow, on which color sector is the spinner LEAST likely to stop?
a)

Blue

b)

Purple

c)

Orange

d)

Red
251

14.

The smaller box contains 20 tickets numbered from 1 to 20. The larger box contains 100
tickets numbered from 1 to 100.
Figure B. 8
Post-Test Item 14 – Ticket Boxes

20 tickets

100 tickets

Without looking at them you can pick a ticket from either box. Which box would give
you the greater chance of picking out a ticket with the number 17 on it?
a)

The box with 20 tickets.

b)

The box with 100 tickets.

c)

Both boxes would give the same chance.

d)

It is impossible to tell.

252

15. A restaurant manager is considering a new location for her restaurant. She anticipates that
the annual cash flow for the new location is:
Annual Cash Flow:
Probability:

30,000

50,000

60,000

90,000

100,000

0.05

0.15

0.30

0.40

???

The expected cash flow for the new location is:
a) 60,000
b) 73,000
c) 66,000
d) 14,600

253

16. An increasing number of employees are exploring the Internet for savings in business
travel. A recent article reported on the results of a survey of 502 corporate travel
managers. Suppose that a contingency table of whether employees research airline ticket
prices and buy airline tickets on the internet revealed the following results:
Table B.8
Post-Test Item 16 Cross Tabulation

Research airline ticket
prices on the internet
Yes
No
Total

Buy airline tickets on the internet
Yes
No

Total

138
52
190

302
198
500

164
146
310

a) Give an example of a simple event.
b) Give an example of a joint event.
c) Determine the probability that a manager selected at random:
i. Buys airline tickets on the internet.
ii. Researches airline ticket prices on the internet.
iii. Researches airline ticket prices on the internet and buys airline tickets on
the internet.
iv. Researches airline ticket prices on the internet or buys airline tickets on
the internet.
d) Calculate the probability that a randomly selected manager researches airline
ticket prices on the internet given that he/she buys airline tickets on the internet.
e) Calculate the probability that a randomly selected manager buys airline tickets on
the internet given that he/she researches airline ticket prices on the internet.
254

f) Determine whether researching airline ticket prices on the internet and buying
airline tickets on the internet are independent.

17. The Cyprus Postal Service claims that it delivers packages on time 85% of the time.
Consider a sample of 20 packages that need to be delivered by the Cyprus Postal Service.
a) Compute the probability that all of these 20 packages are delivered on time.
b) Compute the probability that at least 18 of these packages are delivered on time.
c) Compute the mean and standard deviation of this distribution.

255

APPENDIX C
Instructional Materials
C.1 ACTIVITY 1 (Treatment Group)

Activity Objectives:
i. Students make and justify predictions based on experimental and theoretical probabilities.
ii. Students conduct an experiment to determine experimental probabilities (relative
frequencies).
iii. Students determine theoretical probabilities.
iv. Students compare experimental and theoretical probabilities.
v. Group experimental results are combined to illustrate the Law of Large Numbers.
-------------------------------------------------------------------------------------------------------------------Source:
McConnell, J. W., Brown, S., Usiskin, Z., Senk, S. L., Widerski, T., Anderson, S., Eddins, S.,
Feldman, C. H., Flanders, J., Hackworth, M., Hirschhorn, D., Polonsky, L., Sachs, L., and
Woodward, E. (1998). The University of Chicago School Mathematics Project: Algebra:
Integrated Mathematics (p. 368-369). Glenview, IL: Scott, Foresman and Company.
--------------------------------------------------------------------------------------------------------------------

256

Activity 1: Relative Frequencies, Theoretical Probabilities & the Law of Large Numbers
Exploring the Sum of Two Dice
Instructions:
i. Please complete the following activity in groups as assigned by your instructor. Try to get
everyone in your group involved as you work on the activity.
ii. Please look at the attached note regarding the role that each group member is assigned for the
purposes of this activity. As you work on the activity, remember to carry out your role.
iii. Materials needed: Two six-sided dice and graph paper.
iv. If you have any questions regarding any part of the activity, you may ask any member of
your group or your instructor for help.
v. Please keep the recorder ON at all times; speak clearly and close to the recorder.
vi. Once you finish, make sure you hand in your responses (including this handout and the graph
paper) and the dice to your instructor. You only need to hand in one copy of your responses
for the entire group.
--------------------------------------------------------------------------------------------------------------------Names of Group Members (First and Last Name):

_______________________________________

_______________________________________

_______________________________________

257

Read the instructions carefully as you work on the activity.
1. When two dice are rolled and the numbers on top are added, the sum can be any whole
number from 2 to 12.
i. Which sum do you believe is most likely to occur (i.e. would come up more often if
you roll two dice)?

ii. Why did you pick this number?

2. Your group’s roller should roll the two dice you were given 50 times (Don’t worry; it
sounds a lot but it really isn’t!!). The recorder should record the result in the column
labeled “Frequency” in the table below. The checker should check that the result recorded
is correct. At this time, only fill in Column 2 of the table below.

258

Table C.1
Activity 1 – Recording Results of Sum of Two Dice
Column 1
Sum

Column 2
Frequency

Column 3
My Group’s
Relative Frequency
Fraction
Decimal

Column 4
Probability
(Fraction, decimal or
percentage)

Column 5
Class
Relative
Frequency

2
3
4
5
6
7
8
9
10
11
12

3. Construct a frequency histogram using your group’s results from the 50 dice rolls as you
indicated them in the table above. Use the graph paper provided or the space below.

259

4. After you have rolled the dice 50 times, calculate your group’s relative frequency
(experimental probability) of getting each sum. Write each relative frequency as both a
fraction and a decimal in Column 3 in the table on the previous page.

5. Once you are done with 4 above, have one of your group members go to the board and
record your group’s relative frequencies in the column corresponding to your group’s
number.

6. Based on your table on page 2, how, if at all, would you revise your prediction in part 1?
What patterns do you notice?

7. The diagram below shows the 36 possible outcomes when two dice are rolled.
Figure C.1
Activity 1 – Sample Space for Rolling Two Fair Dice

260

Notice that there is only 1 way to get a sum of 2. We say that the theoretical probability
of getting a sum of 2 is 1/36 ≈ 0,028. In contrast, there are 3 ways of getting a sum of 10
(6 and 4, 5 and 5, 4 and 6), so the probability of getting a sum of 10 is 3/36 ≈ 0,083.
i. Calculate the probability of getting each of the other sums from 2 to 12, and record
these numbers (as a fraction, decimal or percentage) in Column 4 labeled
“Probability” in part 2.
ii. How close are your group’s relative frequencies (experimental probabilities) to the
theoretical probabilities you just computed? i.e. how close are your answers in
Column 3 to your answers in Column 4 on the table in part 2?

8. i. Combine your results with the results of the other groups in your class which are
written on the board. Calculate the relative frequency (experimental probability) for each
sum from 2 to 12 for the entire class. Fill in Column 5 labeled “Class relative frequency”
on the table in part 2.

ii.Which set of relative frequencies – those from your small group, or those from the
whole class – are closer to the theoretical probabilities? i.e. Which are closer to
the results in Column 4: the results in Column 3 or the results in Column 5?

iii.Use the combined results of all the groups in the class to create another
frequency histogram. You may use the graph paper provided or the space below.
261

iv. Compare the shape of this histogram to the histogram you constructed in #3 in part 2.
Is one graph more symmetrical than the other?
If yes, which one is more symmetrical: the histogram for your group’s data or
the histogram for the whole class data?

262

C.2 ACTIVITY 2 (Treatment Group)

Activity Objectives:
i. Conduct an experiment to determine experimental probabilities (relative frequencies).
ii. Construct frequency histograms.
iii. Students compare experimental and theoretical conditional probabilities.
iv. Group experimental results are combined to illustrate the Law of Large Numbers.
v. Explore a situation that relates to conditional probability and independence of events.
vi. Determine sample spaces.
vii. Find the probability of simple events and impossible events.
viii. Apply the multiplication principle for independent events to a sample space.
-------------------------------------------------------------------------------------------------------------------Source: Irini Papaieronymou
-------------------------------------------------------------------------------------------------------------------

263

Activity 2 – Conditional Probability and Independence
Exploring the Sum of Three Dice
Instructions:
i. Please complete the following activity in groups as assigned by your instructor.
ii. Materials needed: Three six-sided dice of different colors.
iii. If you have any questions regarding any part of the activity, you may ask any member of
your group or your instructor for help.
iv. Respond to all the questions asked on this handout in the best way you can and use this
handout to fill in your responses.
v. Please keep the recorder ON at all times; speak clearly and close to the recorder.
--------------------------------------------------------------------------------------------------------------------Names of Group Members and Role:

(roller – rolls the dice) _______________________________________

(recorder – writes down the answers) _______________________________________

(checker – checks that answers are recorded correctly ) ______________________________

Today you will explore the sum of three dice by rolling a set of three fair six-sided dice.
You will first perform an experiment and then will use your results to answer questions
regarding conditional probability and independence. We will study conditional probability and
independence in more detail next time in class.
264

1. Roller:

Roll the three dice simultaneously (at the same time) 50 times.

Recorder: Record the result showing on each die and the sum of the three dice on
each roll in the table below.
Checker: Check that the answers recorded are correct.
Table C.2
Activity 2 – Results of Rolling Three Fair Dice
Roll
Red Die

Result
Green Die

Sum of the three dice
Purple Die

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
265

Table C.2 (cont’d)
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

266

2.

Go back to your results of the set of 50 three-dice rolls in the table on the previous page.

a. Based on your results, what is the relative frequency that one of the dice comes up 3?

b. Based on your results, if the sum is 6, what is the relative frequency that one of the dice is
3?

c. Based on your answers in 2a and 2b, would you say that the relative frequency of one of
the dice coming up 3 is affected by knowledge that the sum is 6?
(Hint: Is your answer in 2a equal to your answer in 2b?)

267

Note: The relative frequency you have computed in part 2b is based on the condition that one of
the dice resulted in 3. Relative frequencies that are based on certain pre-set conditions can also
be determined theoretically using conditional probability.

Let us begin to determine the relative frequency computed in part 2 theoretically.

3.a.

How many outcomes are possible when three dice are rolled?

b.

What are the different ways in which three dice can give you a sum of 6?

c.What is the probability that you will get a sum of 6 when three dice are rolled?

268

4. Theoretically, the conditional probability of an event A given an event B is as follows:

P(event A given event B) = P(A and B)
P(B)

or symbolically, P( A | B) 

P( A  B)
(Formula 2)
P( B)

a. Use Formula 2 above to find the theoretical probability that, when three fair dice are
rolled, one of the dice comes up 3 given that the sum is 6.

Event A =

Event B =

P(A | B) =

269

b. How close is your answer in 4a to your answer in 2b?

In addition to computing conditional probabilities theoretically, we can also use probability
theory to determine whether the probability of one event affects the probability of another event.

If the probability of one event affects the probability of another event, then the two events are
said to be dependent.
If the probability of one event does not affect the probability of another event, then the two
events are said to be independent.

Theoretically, statistical independence of two events can be computed as follows:

Let A and B be two events. Then, A and B are independent if and only if P(A|B) = P(A)
(Formula 3)

270

With regards to this activity:

Let Event A = one of the dice comes up 3 and Event B = get a sum of 6

5.a.Use Formula 3 on the previous page to determine whether Events A and B are independent.
Show your work.

b. Does your conclusion in 5a agree with your answer to 2c?

271

C.3 ACTIVITY 3 (Treatment Group)
Activity Objectives:
i. Students make and justify predictions based on experimental and theoretical probabilities.
ii. Help students overcome probabilistic misconceptions: equiprobability and outcome
orientation.
iii. Students study the probability distribution of a random variable.
iv. Students study sample spaces.
--------------------------------------------------------------------------------------------------------------------Source: Khazanov, L. (2008). Addressing students’ misconceptions about probability during
the first years of college. Mathematics and Computer Education, 42(3), 180-192.
The following is the original activity as described by Khazanov:
1. Instructor explains the rules of the game. Numbers 1 through 15 are written on a number line.
Participants have 36 (or 48) chips to place above these numbers. The instructor or a
designated student repeatedly rolls two dice. The sum of the numbers showing is calculated.
If a student has a chip above the number on the line that matches the sum of the numbers on
the dice, then she removes any chips. The winner is the person or the group that has all their
chips removed ahead of any other group or person.
2. Once the game is over, engage students in a discussion. Ask them to provide a rationale for
their distributions. Then ask which distribution won, and why. Some students could have
chosen a uniform distribution from 2 to 12 (equiprobability bias), others might have placed
all of their chips above 7 (outcome orientation misconception). It is important for these
students to understand why they made poor choices. However, it is equally important to
emphasize that some of the students who made poor choices stood a chance winning. They
272

were not destined to lose, but had a smaller chance of winning than those who placed more
chips towards the center of the distribution. Students’ understanding may be reinforced by
constructing a 6*6 table listing 36 equally likely outcomes when two dice are tossed. Ask
your students to compute the sum for each entry and observe that while the possible values
for the sums are 2 to 12, the number of times these numbers occur is different.
3. A good additional question: what mistake could one make to totally eliminate one’s chance
of winning (make the probability of winning equal to 0)? The mistake, of course, is placing a
chip over 1 or any number greater than 12.
--------------------------------------------------------------------------------------------------------------------

273

Activity 3: Discrete Probability Distributions
A Game of Chance: Where Should The Chips Be Placed?

Instructions:
i. Please complete the following activity in groups as assigned by your instructor.
ii. Materials needed: Two dice of different colors, 36 chips of three different colors, and a
poster on which a number line is drawn.
iii. Remember to keep the recorder on during the entire activity and to talk close to it.
iv. If you have any questions regarding any part of the activity, you may ask any member of
your group or your instructor for help.
v. Please respond to all the questions asked on this handout in the best way you can and use this
handout to fill in your responses.
--------------------------------------------------------------------------------------------------------------------Names of Group Members:
(roller: rolls the dice) _______________________________________
(recorder: writes down the question) _______________________________________
(checker: checks that the answers are correct) _____________________________________

Today you will play a game that relates to the sum of two dice. In this game, you will each pick a
color of chips (12 chips for each person), place your chips over any numbers on the number line
provided on the poster, throw a pair of dice, and after each roll, remove any chips you have over
the number that corresponds to the sum of the two dice. The group member who has his/her
chips removed first wins the game.
274

1. In your groups, pick one of the available colors of chips and gather all 12 chips of that
color so that you can use them in this activity. Each group member should have 12 chips
of the same color.
2. On the separate poster paper given to you by your instructor, a number line is provided
with the numbers 1 through 15 written on it.
3. Place your 12 chips above any of the numbers on the number line. Distribute all of your
chips in any way you want on the number line. You may place all of them over one
number or you may spread them out over various numbers.
4. Before you start playing the game, fill in the table in part 5 and respond to question 6.

275

5. In the following table, indicate how many chips each member of your group has placed
above each of the numbers on the number line.
Table C.3
Activity 3 – Placement of Chips on Number Line
Number on the
Number Line

Student Name
Student Name
Student Name
_____________________ ______________________ _____________________

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

6. Provide an explanation for the way that each group member placed their chips on the
number line.
Student Name and Explanation:

Student Name and Explanation:
276

Student Name and Explanation:

*Before you begin to play the game, read question 7 carefully.
7. Roller: Roll the two dice repeatedly at the same time and call out their sum so that the
recorder can record this down on the table provided below.
Recorder: In the table below, indicate the sum of the numbers on the dice on each
roll.
Checker: Check that the results recorded are correct.

As the two dice are rolled, if you have any chips above the number on the line that
matches the sum of the numbers on the two dice just rolled, then remove one of these
chips. Continue rolling the dice, each time removing chips as indicated above and
recording the sum in the table below. You may add more rows to the table below if
necessary. Stop filling in the table below once a group member has all of his/her chips
removed that is, once you have a winner!

277

Table C.4
Activity 3 – Recording Result of Rolling Two Fair Dice
Roll
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Sum

278

8. How many times did you roll the dice before you had a winner in your group? That is, on
which roll did someone in your group have all of his/her chips removed? _____

9. Who won the game in your group? _________________________

10. Why do you think this group member won the game?

11. Were the other two group members destined to lose? That is, did they have NO chance of
winning? Explain.

12. What mistake could you make to totally eliminate your chance of winning? That is, what
mistake could you do to make your probability of winning equal to 0?

279

For the following questions on this activity, use the diagram provided below which shows the
possible outcomes when two dice are rolled.
Figure C.2
Activity 3 – Sample Space for Rolling Two Fair Dice

13. How many possible outcomes are there when rolling two fair dice? _______

14. Which are the possible sums when two fair dice are rolled?

15. How is the number line connected to the sums you listed in question 14) above?

280

16. Given the list of possible sums that you have identified in 14) above, which numbers on
the number line would eliminate one’s chance of winning? i.e. which numbers would
make the chance of winning 0?

17. In the table below, the left column indicates the possible sums when rolling two fair dice.
Fill in the right column in the table, to indicate the theoretical probability for each of
these sums. The two columns of this table taken together make up the discrete probability
distribution for the sum of two fair dice. We will study discrete probability distributions
next time in class (chapter 5)
Table C.5
Activity 3 – Probability Distribution for Sum of Two Dice
Sum of the numbers on two fair dice
2
3
4
5
6
7
8
9
10
11
12

Probability

18. What is the sum of the values of the probability column in the table above?

19. Based on the theoretical probabilities that you have computed in 17) above, which sums
give the best chance of winning the game you have just played in your groups?
281

C.4 ACTIVITY 4 (Treatment Group)
Activity Objectives:
i. Explore a binomial distribution problem.
ii. pproach a binomial probability problem through an investigation of relative frequencies
and through an analysis of outcomes.
iii. Compare the experimental to the theoretical approach of solving a binomial distribution
problem.
--------------------------------------------------------------------------------------------------------------------Source: Shaughnessy, M. J., Barrett, G., Billstein, R., Kranendonk, H. A., and Peck, R. (2004).
What’s the probability of a hit? In Lott, J. W., and House, P. A. (Eds.), Navigating through
probability in grades 9-12, p. 40-42 and p. 94-97. Reston, VA: The National Council of Teachers
of Mathematics, Inc.
Modifications done to original activity:
i) Context was changed. The original activity concerned a baseball player. Since baseball is not
popular in Cyprus, the context was changed to be about a football player (football is the most
popular type of sports game in Cyprus). So, the activity is about football goals not baseball hits,
as was the case in the original.
ii) In the original activity, the baseball player had a batting average of 0.4. In the modified
activity, the football player has a scoring average of 1/6. This was changed so that in later parts
of the activity, the students can use a die to generate data.
iii) In the original activity, the students were asked to perform two sets of simulations: a set of 20
simulations and then a set of another 50 simulations. Due to time constraints, I believe that
performing both of these sets of simulations would make the activity last longer than two 50
282

minute class periods but the goal is for all activities used to take up to two periods. So, in the
modified version of the activity, the students are asked to do a set of 50 simulations only in their
groups. After that, they will use the results of all groups which the instructor will record on the
board instead of performing more simulations in their groups.
iv) Part 5)b) was added.
--------------------------------------------------------------------------------------------------------------------

283

Activity4 – Binomial Distribution
What’s the Probability of Making a Basket?
Instructions:
i. Please complete the following activity in groups as assigned by your instructor.
ii. Materials needed: a table of random numbers.
iii. Remember to keep the recorder on at all times (do NOT pause or stop it during the
activity). Speak clearly and close to the recorder.
iv. If you have any questions regarding any part of the activity, you may ask any member of
your group or your instructor for help.
v. Please respond to all the questions asked on this handout in the best way you can and use
this handout to fill in your responses.
--------------------------------------------------------------------------------------------------------------------Names of Group Members:

(recorder: writes down the answers) _______________________________________

(asks questions) _______________________________________

(checker) _______________________________________

284

Suppose that a basketball player has a free-throw average of 70%. Thus, any time he attempts a
free throw during a basketball game, this player has a constant probability of making the basket,
or P(B), of 70% and a constant probability of not making the basket, or P(N), of 30%.

1)a) Let B = event that the player makes the basket
Let N = event that the player does not make the basket
Using the letters B and N construct the sample space (the set of all possible outcomes/
combinations) for three free-throw attempts.

b) Which outcome(s) in part 1a include at least 2 baskets?

285

2)a) Compute the probabilities (as a percentage) of each of the individual outcomes in your
sample space in part 1a. Remember: the player’s probability of making the basket is
P(B) = 70% and his probability of not making the basket is P(N) = 30%.

2b) Use the probabilities in 2a above to calculate the theoretical probability that the basketball
player would make at least 2 baskets in three free-throw attempts.

286

3) a) Use the table of random numbers given to you by your instructor to indicate data on 50 sets
of three free-throw attempts for the player with a 70% free-throw average. Fill in the table below.
Table C.6
Activity 4 – Results of Free-Throw Basketball Attempts
Outcome
Number of baskets made
(three free-throw attempt)

287

Table C.6 (cont’d)

3)b) On the basis of your 50 sets of simulated data for three free-throw attempts, what is the
probability that the player will make at least 2 baskets in three free-throw attempts?

c) How does your result from 3)b) compare with your calculation in question 2b?

288

4) Turn to one of the other groups of students close to your group.

Name the students in that group:
_______________________________________________

a) Compare the results of your simulations of 50 sets of three free-throw attempts with
those of the other group. How close is your probability for at least 2 baskets in three
free-throw attempts to the probability of the other group?

b) If the probabilities differ why do you think this is the case?

289

5) On the board, your instructor has recorded the results of at least two baskets in 50 sets of
three free-throw attempts of all groups. Your group’s results have also been recorded.
Combine all class results to answer the following:
a) Based on the results from all groups, what is the probability that the player makes at
least 2 baskets in three free-throw attempts?

b) How does your calculation in 5a) compare with your calculation in question 2b?

c) How does your answer to 5a) compare with your group’s simulated results of 50 sets
of three free-throw attempts?

290

The probabilities that you have found in this activity can also be derived using probability
theory. The ideas behind this activity are related to what is called a binomial probability
distribution. The binomial probability distribution involves the following formula:

P(X  x) =

n x
   p  (1  p) n  x
 x
 

where n is the number of observations, p is the probability of success (i.e. making the basket)
and x is the number of baskets made.

So, in this activity, theoretically, the probability of the player making zero baskets in three
free-throw attempts is:
0

3

3

3  4   2 
8
1
2

P( X = 0) =          1  1    
 0 6
216 27
6
    6

6) Using the formula for the binomial distribution above:
a) Compute the probability that the player makes 2 baskets in three free-throw attempts.

b) Compute the probability that the player makes 3 baskets in three free-throw attempts.

291

c) Compute the probability that the player makes at least 2 baskets in three free-throw
attempts.

292

C. 5 IN-CLASS PROBLEMS ASSIGNED TO CONTROL GROUP / GROUPWORK

Problems From the Course Textbook:
Levine, D. M., Krehbiel, T. C., and Berenson, M. L. (2010). Business statistics: A first course
(5

th

ed.). Upper Saddle River, NJ: Pearson Education, Inc.

p. 157-158

4.2 An urn contains 10 red balls and 8 green balls. One ball is to be selected from the urn.
a. Give an example of a simple event.
b. What is the complement of a red ball?

4.8 According to an Ipsos poll, the perception of unfairness in the U.S. tax code is spread fairly
evenly across income groups, age groups, and education levels. In an April 2006 survey of 1,005
adults, Ipsos reported that almost 60% of all people said the code is unfair, whereas slightly more
than 60% of those making more than $50,000 viewed the code as unfair (“People Cry
Unfairness”, The Cincinnati Enquirer, April 16, 2006, p. A8). Suppose that the following
contingency table represents the specific breakdown of responses:
Table C.7
Problem 4.8 Cross Tabulation

U.S. Tax Code
Fair
Unfair
Total

Income Level
Less than $50,000 More than $50,000
225
180
280
320
505
500
293

Total
405
600
1,005

a)

Give an example of a simple event.

b)

Give an example of a joint event.

c)

What is the complement of “tax code is fair”?

d)

Why is “tax code is fair and makes less than $50,000” a joint event?

4.9 Referring to the contingency table is Problem 4.8, if a respondent is selected at random, what
is the probability that he or she
a. thinks the tax code is unfair?
b. thinks the tax code is unfair and makes more than $50,000?
c. thinks the tax code is unfair or makes more than $50,000?
d. Explain the difference in the results in (b) and (c).

p. 165
4.23 According to an Ipsos poll, the perception of unfairness in the U.S. tax code is spread fairly
evenly across income groups, age groups, and education levels. In an April 2006 survey of 1,005
adults, Ipsos reported that almost 60% of all people said the code is unfair, whereas slightly more
than 60% of those making more than $50,000 viewed the code as unfair (“People Cry
Unfairness”, The Cincinnati Enquirer, April 16, 2006, p. A8). Suppose that the following
contingency table represents the specific breakdown of responses:

294

Table C.8
Problem 4.23 Cross Tabulation

U.S. Tax Code
Fair
Unfair
Total

Income Level
Less than $50,000 More than $50,000
225
180
280
320
505
500

Total
405
600
1,005

a. Given that a respondent earns less than $50,000, what is the probability that he or
she said that the tax code is fair?
b. Given that a respondent earns more than $50,000, what is the probability that he
or she said that the tax code is fair?
c. Is income level independent of attitude about whether the tax code is fair?
Explain.

295

p. 183
5.3 The manager of a large computer network has developed the following probability
distribution of the number of interruptions per day:
Table C.9
Problem 5.3 Probability Distribution
Interruptions (X)
0
1
2
3
4
5
6

P(X)
0.31
0.34
0.18
0.10
0.04
0.02
0.01

Compute the expected number of interruptions per day.

5.4 In the carnival game Under-or-Over-Seven, a pair of fair dice is rolled once, and the resulting
sum determines whether the player wins or loses his or her bet. For example, the player can bet
$1 that the sum will be under 7 – that is, 2, 3, 4, 5, or 6. For this bet, the player wins $1 if the
result is under 7 and loses $1 if the outcome equals or is greater than 7. Similarly, the player can
bet $1 that the sum will be over 7 – that is, 8, 9, 10, 11 or 12. Here, the player wins $1 if the
result is over 7 but loses $1 if the result is 7 or under. A third method of play is to bet $1 on the
outcome 7. For this bet, the player wins $4 if the result of the roll is 7 and loses $1 otherwise.
a. Construct the probability distribution representing the different outcomes that
are possible for a $1 bet on under 7.
b. Construct the probability distribution representing the different outcomes that
are possible for a $1 bet on over 7.
296

c. Construct the probability distribution representing the different outcomes that
are possible for a $1 bet on 7.
d. Show that the expected long-run profit (or loss) to the player is the same, no matter
which method of play is used.

p. 190
5.13 When a customer places an order with Rudy’s On-Line Office Supplies, a computerized
accounting information system (AIS) automatically checks to see if the customer has exceeded
his or her credit limit. Past records indicate that the probability of customers exceeding their
credit limit is 0.05. Suppose that, on a given day, 20 customers place orders. Assume that the
number of customers that the AIS detects as having exceeded their credit limit is distributed as a
binomial random variable.
a. What are the mean and standard deviation of the number of customers
exceeding their credit limits?
b. What is the probability that 0 customers will exceed their limits?
c. What is the probability that 1 customer will exceed his or her limit?
d. What is the probability that 2 or more customers will exceed their limits?

297

APPENDIX D
SCORING RUBRICS FOR POST-TEST OPEN-ENDED ITEMS

Table D.1
Scoring Rubric for Post-Test Item 16a
Code/Score Response
1
Correct Response:
Student specifies one of the following:
i)Employee buys airline tickets on the internet
ii)Employee does not buy airline tickets on the internet
iii)Employee researchers airline ticket prices on the internet
iv) Employee does not research airline ticket prices on the internet
0

999

Incorrect Response:
Student provides a response different from any of the four correct
statements provided above.
Non-response:
Blank or incomplete sentence provided.

298

Table D.2
Scoring Rubric for Post-Test Item 16b
Code/Score Response
1
Correct Response:
Student specifies one of the following:
i)Employee researches airline ticket prices on the internet and buys
airline tickets on the internet.
ii) Employee researches airline ticket prices on the internet and does not
buy airline tickets on the internet.
iii) Employee does not research airline ticket prices on the internet and
buys airline tickets on the internet.
Iv)Employee does not research airline ticket prices on the internet and
does not buy airline tickets on the internet.
0

999

Incorrect Response:
Student provides a response different from any of the four correct
statements provided above.
Non-response:
Blank or incomplete sentence provided.

299

Table D.3
Scoring Rubric for Post-Test Items 16c) i), 16c) ii) and 16c) iii)
Code/Score Response
2
Correct Response:
For part i) 190/500 or 19/50 or 0.38 or 38%
For part ii) 302/500 or 0.604 or 60.4%
For part iii) 138/500 or 0.276 or 27.6%
1

Partially Correct Response:
Either the numerator or the denominator in the fraction provided is
correct (but not both).

0

Incorrect Response:
Neither the numerator nor the denominator in the fraction provided is
correct or an incorrect decimal or percent is provided.

999

Non-response:
Blank

300

Table D.4
Scoring Rubric for Post-Test Item 16c) iv)
Code/Score Response
2
Correct Response:
Student correctly uses the union rule for finding the probability of events
and arrives at the correct answer:
302/500 + 190/500 – 138/500 = 352/500 or 0.704 or 70.4%
1

Partially Correct Response:
Student either
i)states the union rule correctly but does not carry out the computations
or
ii) states the union rule correctly but some computations are incorrect
or
iii)specifies the correct computation to be carried out
i.e. 302/500 + 190/500 – 138/500 but does not provide a final answer
or
iv)provides a correct final answer but does not show any computations or
v)provides incorrect intermediate computations

0

999

Incorrect Response:
The answer provided is completely incorrect.
Non-response:
Blank

301

Table D.5
Scoring Rubric for Post-Test Items 16d and 16e
Code/Score Response
2
Correct Response:
Student correctly identifies a conditional probability:
For part d: 138/190
For part e: 138/302
The student may have used the rule for conditional probability correctly
or might have correctly used the necessary cells from the table.
1

Partially Correct Response:
Student either
i)indicates that a conditional probability should be used and/or states the
rule for conditional probability correctly but does not carry out the
computations
or
ii)specifies the correct computation to be carried out
i.e. (138/500) / (190/500) but does not provide a final answer
or
iii)provides a correct final answer but does not show any computations or
provides incorrect intermediate computations
or
iv)provides the reverse fraction i.e. 190/138 (part d) or 302/138 (part e)
or
v) provides a fraction in which either the numerator or the denominator is
correct (but not both)

0

999

Incorrect Response:
The answer provided is completely incorrect.
Non-response:
Blank

302

Table D.6
Scoring Rubric for Post-test Item 16f
Code/Score Response
2
Correct Response:
Student correctly
i)indicates that the two events are not independent:
and
ii) shows this through computations that make use of conditional
probability
1

Partially Correct Response:
Student
i)states that the two events are not independent but does not show any
computations or provides incorrect computations or does not provide
any reasoning for this conclusion
or
ii) student carries out correct computations that relate to the
independence of events but does not indicate whether the events are
independent or not
or
iii) student carries out correct computations that relate to the
independence of events but incorrectly indicates that the two events are
independent

0

999

Incorrect Response:
Student indicates that the two events are independent and/or provides
incorrect computations/reasoning when arriving at this conclusion.
Non-response:
Blank

303

Table D.7
Scoring Rubric for Post-Test Item 17a
Code/Score
2
Correct Response:

Response

0.85 20  0.039  3.9%

1

Partially Correct Response:
Either the base or the exponent is incorrect
Or
Student provides correct computations but incorrect final answer

0

Incorrect Response:
Incorrect computations and incorrect final answer.

999

Non-response:
Blank

304

Table D.8
Scoring Rubric for Post-Test Item 17b
Code/Score
Response
2
Correct Response:
i)Student recognizes that the binomial distribution should be used and
that P( X  18) should be computed.
and
ii)Student carries out the computations correctly using the formula for
finding the probability of a binomial distribution
and
iii)Student arrives at the correct final answer of 40.6%
1

Partially Correct Response:
i)Student recognizes that the binomial distribution should be used and
that P( X  18) should be computed but does not carry out any
computations or the computations are incorrect
or
ii)Student carries out the computations correctly using the formula for
finding the probability of a binomial distribution but does not arrive
at a final answer
or
iii)Student only provides the final answer of 40.6%
or
iv)Student computes P(X = 18) or P(X ≤ 18)

0

Incorrect Response:
Incorrect computations and incorrect final answer.

999

Non-response
Blank

305

Table D.9
Scoring Rubric for Post-Test Item 17c
Code/Score
Response
2
Correct Response:
Student correctly computes
i) The mean i.e. 20 x 0.85 = 17
and
ii) The variance i.e. 20 x 0.85 x 0.15 = 2.55
1

Partially Correct Response:
i)Student computes either the mean or the variance correctly but not
both.
or
ii)Student shows the computations for the mean and the variance but
does not provide a final answer for either
or
iii)Student provides a correct final answer for both the mean and the
variance but does not show any computations.

0

Incorrect Response:
Incorrect computations and incorrect final answer.

999

Non-response
Blank

306

Table D.10
Scoring Rubric for Post-Test Item 18a
Code/Score
Response
1
Correct Response:
Student provides one of the following simple events:
i) Textbook published is a success
ii) Textbook published is break-even
iii) Textbook published is a loser
iv) Textbook received favorable reviews
0

999

Incorrect Response:
Incorrect simple event provided.
Non-response:
Blank

307

Table D.11
Scoring Rubric for Post-Test Items 18b, 18c, and 18d
Code/Score
Response
2
Correct Response:
Student recognizes that conditional probability should be used and
provides the correct answer:
For part b: 80%
For part c: 60%
For part d: 20%
1

Partially Correct Response:
Student recognizes that conditional probability should be used but does
not provide a correct numerical final answer.

0

Incorrect Response:
Student does not recognize that conditional probability should be used
and does not provide a correct answer.

999

Non-response:
Blank

308

Table D.12
Scoring Rubric for Post-Test Item 18e
Code/Score
Response
2
Correct Response:
Student recognizes that Bayes’ Theorem should be used and correctly
applies the theorem to arrive at the correct final answer:
P(Success|Favorable Reviews) =
(0.8x0.3)/[(0.8x0.3)+(0.6x0.5)+(0.2x0.2)] = 0.24 / 0.58 = 0.414
1

Partially Correct Response:
Student recognizes that Bayes’ Theorem should be used and attempts to
use it but not all computations are correct
Or
Student recognizes that Bayes’ Theorem should be used, correctly uses it
but arrives at an incorrect final answer.

0

Incorrect Response:
Student recognizes that Bayes’ Theorem should be used but computations
and final answer are incorrect.

999

Non-response:
Blank

309

APPENDIX E: STATISTICAL RESULTS

Table E.1
Percent-Correct Responses on Pre-Test and Post-Test Multiple-Choice Items

Item
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Control
(Group B; morning)
N = 17
PreTest PostTest
94.1
76.5
70.6
70.6
70.6
64.7
100
47.1
23.5
29.4
23.5
11.8
11.8
17.6
17.6
5.9
52.9
52.9
70.6
76.5
94.1
35.3
58.8
52.9
88.2
47.1
88.2
58.8
23.5

Percent-Correct Responses
Treatment1
Treatment2
(Group A; morning) (Group C; evening)
N= 20
N=7
PreTest PostTest PreTest PostTest
80
85
85.7
100
95
85
85.7
71.4
75
90
42.9
100
90
80
85.7
85.7
30
55
28.6
42.9
45
30
14.3
42.9
50
55
14.3
71.4
15
15
14.3
14.3
85
70
42.9
71.4
40
95
28.6
57.1
95
90
57.1
85.7
95
95
85.7
85.7
80
75
71.4
71.4
80
75
71.4
85.7
40
71.4

310

Treatment1
& Treatment2
N = 27
PreTest PostTest
81.5
88.9
92.6
81.5
66.7
92.6
88.9
81.5
29.6
51.9
37
33.3
40.7
59.3
14.8
14.8
74.1
70.4
37
85.2
85.2
88.9
92.6
92.6
77.8
74.1
77.8
77.8
48.1

Table E.2
Results of Distractor Analysis of Multiple-Choice Items (Correct Response in bold)

Item
1

2

3

4

5

6

7

8

9

10

Distractor
Or
Correct
Choice
A
B
C
A
B
C
A
B
C
A
B
C
D
A
B
C
D
A
B
C
A
B
C
D
E
A
B
C
A
B
C
A
B
C
D

Control
(Group B; morning)
N = 17
Pre-Test (%) Post-Test (%)
5.9
0
94.1
23.5
5.9
70.6
17.6
70.6
11.8
0
0
0
100
52.9
23.5
0
23.5
58.8
23.5
17.6
11.8
29.4
11.8
11.8
23.5
64.7
17.6
11.8
23.5
17.6
52.9
5.9
5.9
70.6
0

17.6
0
76.5
11.8
11.8
70.6
11.8
64.7
17.6
11.8
0
41.2
47.1
29.4
29.4
23.5
17.6
35.3
11.8
47.1
29.4
29.4
0
17.6
17.6
82.4
5.9
5.9
23.5
17.6
52.9
11.8
5.9
76.5
0
311

Treatment1 & Treatment2
N = 27
Pre-Test (%) Post-Test (%)
3.7
14.8
81.5
7.4
0
92.6
18.5
66.7
7.4
0
0
7.4
88.9
44.4
29.6
11.1
7.4
44.4
37
14.8
7.4
14.8
0
40.7
18.5
70.4
14.8
11.1
11.1
11.1
74.1
0
0
37
11.1

3.7
7.4
88.9
3.7
7.4
81.5
0
92.6
3.7
3.7
0
14.8
81.5
29.6
51.9
3.7
14.8
51.9
33.3
11.1
11.1
7.4
0
59.3
22.2
81.5
14.8
3.7
11.1
14.8
70.4
7.4
3.7
85.2
3.7

Table E.2 (cont’d)
11

12

13

14

15

A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D

0
5.9
94.1
0
58.8
5.9
29.4
5.9
5.9
88.2
5.9
0
88.2
0
0
5.9

47.1
11.8
35.3
5.9
52.9
11.8
23.5
11.8
0
47.1
35.3
11.8
58.8
5.9
35.3
0
29.4
23.5
11.8
35.3

312

0
11.1
85.2
0
92.6
3.7
0
0
3.7
77.8
14.8
0
77.8
3.7
7.4
7.4

7.4
0
88.9
3.7
92.6
0
3.7
3.7
0
74.1
25.9
0
77.8
0
22.2
0
29.6
48.1
18.5
3.7

BIBLIOGRAPHY

313

BIBLIOGRAPHY

Aczel, A. D. and Sounderpandian, J. (2002). Complete business statistics 5th ed. New York, NY:
McGraw-Hill Companies, Inc.
Albert, J. H. (2003). College students’ conceptions of probability. The American Statistician,
57(1), 37-45.
Aliaga, M., Cobb, G., Cuff, C., Garfield, J., Gould, R., Lock, R., Moore, T., Rossman, A.,
Stephenson, B., Utts, J., Velleman, R., and Witmer, J. (2005). Guidelines for assessment and
instruction in statistics education (GAISE) college report. Retrieved April 27, 2008, from
http://www.amstat.org/education/gaise/
Amit, M. and Jan, I. (2006). Autodidactic learning of probability concepts through games. In
Novotná, J., Moraová, H., Krátká, M. and Stehlíková, N. (Eds.). Proceedings of the 30th
Conference of the International Group for the Psychology of Mathematics Education, 2, 4956. Prague: PME.
Anderson, J. R. (1982). Acquisition of cognitive skill. Psychological Review, 89, 369-406.
Antoine, W. (2000). An exploration of the misconceptions and incorrect strategies liberal arts
students use when studying probability. Doctoral dissertation, Columbia University.
Aquilonius, B. C. (2005). How do college students reason about hypothesis testing in
introductory statistics courses? Doctoral dissertation, University of California, Santa
Barbara, CA.
Aspinwall, L. and Tarr, J. E. (2001). Middle school students’ understanding of the role sample
size plays in experimental probability. Journal of Mathematical Behavior, 20, 229-245.
Baldwin, R. G. (2009). The climate for undergraduate teaching and learning in STEM fields.
New Directions for Teaching and Learning, 117, 9-17.
Bamberger, M. E. (2003). Methods college students use to solve probability problems and the
factors that support or impede their success. Doctoral dissertation, Oregon State University.
Bao, L. (2006). Theoretical comparisons of average normalized gain calculations. American
Journal of Physics, 74(10), 917-922.
Batanero, C., Henry, M., and Parzysz, B. (2005). The nature of chance and probability. In Jones,
G. A. (Ed.), Exploring Probability in School: Challenges for Teaching and Learning (pp.
15-37). New York, NY: Springer Science and Business Media, Inc.
314

Batanero, C. and Sanchez, E. (2005). What is the nature of high school students’ conceptions and
misconceptions about probability? In Jones, G. A. (Ed.), Exploring Probability in School:
Challenges for Teaching and Learning (pp. 241-266). New York, NY: Springer Science and
Business Media, Inc.
Batanero, C., Serrano, L., & Garfied, J. B. (1996). Heuristics and biases in secondary school
students’ reasoning about probability. In L. Puig & A. Gutierrez (Eds.), Proceedings of the
20th conference of the International Group for the Psychology of Mathematics Education, 2,
51-58. Valencia, Spain: University of Valencia.
Batt, J. (2004). Stollen innocence. A mother’s fight for justice: The authorized story of Sally
Clark. Ebury Press.
Benko, P. (2006). Study of the development of students’ ideas in probability. Doctoral
dissertation, Rutgers, The State University of New Jersey.
Burdzy, K. (2009). The search for certainty: On the clash of science and philosophy of
probability. Singapore: World Scientific Publishing Co. Pte. Ltd.
Carter, T. A. (2005). Knowledge and understanding of probability and statistics topics by
preservice PK-8 teachers. Doctoral dissertation, Texas A & M University.
Chance, B. L. and Garfield, J. B. (2002). New approaches to gathering data on student learning
for research in statistics education. Statistics Education Research Journal, 1, 38-44.
Cobb, G. (2000). Teaching statistics: More data, less lecturing. In T. L. Moore (Ed.), Teaching
statistics: Resources for undergraduate instructors. MAA Notes, 52, 3-5. Washington, DC:
Mathematical Association of America Inc.
Cohen, D. (Ed.) (1995). Crossroads in mathematics: Standards for introductory college
mathematics before calculus. Memphis, TN: American Mathematical Association of TwoYear Colleges.
Cole, M., John-Steiner, V., Scribner, S. and Souberman, E. (Eds.) (1978). L.S. Vygotsky; Mind in
society: The development of higher psychological processes. Cambridge, MA: Harvard
University Press.
Cooper, J., and Mueck, R. (1990). Student involvement in learning: Cooperative learning and
college instruction. Journal of Excellence in College Teaching, 1(1), 68-76.
Cyprus University of Technology, 2011. Retrieved October 20, 2011, from http://www.cut.ac.cy/
Dansereau, D. F. (1988). Cooperative learning strategies. In C. E. Weinstein, E. T. Goetz, & P.
A. Alexander, (Eds.), Learning and study strategies: Issues in assessment, instruction, and
evaluation (pp. 103-120). Orlando, FL: Academic Press.
315

Dees, R. L. (1991). The role of cooperative learning in increasing problem-solving ability in a
college remedial course. Journal for Research in Mathematics Education, 22(5), 409-421.
delMas, R. C., Ooms, A., Garfield, J. B., & Chance, B. (2006). Assessing students’ statistical
reasoning. In A. Rossman & B. Chance (Eds.), Working cooperatively in statistics
education: Proceedings of the Seventh International Conference on Teaching Statistics.
Voorburg, The Netherlands: International Statistical Institute. Retrieved July 2, 2009, from
http://www.stat.auckland.ac.nz/~iase/publications/17/6D3_DELM.pdf
Derry, S. J., Levin, J. R., Osana, H. P., Jones, M. S., and Peterson, M. (2000). Fostering students’
statistical and scientific thinking: Lessons learned from an innovative college course.
American Educational Research Journal, 37(3), 747-773.
Dewey, J. (1933). How we think. Lexington, MA: D. C. Heath.
Dewey, J. (1969). Experience and education (11th printing). Toronto, Ontario: Collier-Macmillan
Canada Ltd.
Diaz, C. and de la Fuente, I. (2007). Assessing students’ difficulties with conditional probability
and Bayesian reasoning. International Electronic Journal of Mathematics Education, 2(3),
128-148.
Effandi, Z. and Zanaton, I. (2007). Promoting cooperative learning in science and mathematics
education: A Malaysian perspective. Eurasia Journal of Mathematics, Science and
Technology Education, 3(1), 35-39.
Esmonde, I. (2009). Ideas and identities: Supporting equity in cooperative mathematics learning.
Review of Educational Research, 79(2), 1008-1043.
European University Cyprus. (2007). Programs of study 2007-2008. Nicosia, Cyprus: Author.
Evans, J. and Pollard, P. (1985). Intuitive statistical inferences about normally distributed data.
Acta Psychologica, 60, 57-71.
Even, R. & Tirosh (2008). Teacher knowledge and understanding of students’ mathematical
learning and thinking. In L. D. English (Ed.), Handbook of International Research in
Mathematics Education (2nd Ed.) (pp. 202-222). New
York, NY: Routledge.
Exarchakos, T. G. (1988). Διδακτική των μαθηματικών [Didactics of mathematics]. Athens,
Greece: Ellinika Grammata.
Falk, R. (1986). Misconceptions of statistical significance. Journal of Structural Learning, 9, 8396.
316

Falk, R. (1988). Conditional probabilities: Insights and difficulties. In R. Davidson & J. Swift
(Eds.), The Proceedings of the Second International Conference on Teaching Statistics.
Victoria, B.C.: University of Victoria.
Field, A. (2000). Discovering statistics using SPSS for windows. London, UK: SAGE
Publications Ltd.
Fischbein, E. (1975). The intuitive sources of probabilistic thinking in children. Dordrecht,
Holland: D. Reidel Publishing Company.
Fischbein, E., and Gazit, A. (1984). Does the teaching of probability improve probabilistic
intuitions? Educational Studies in Mathematics, 15, 1-24.
Fischbein, E., Nello, M. S., & Marino, M. S. (1991). Factors affecting probabilistic judgments in
children in adolescence. Educational Studies in Mathematics, 22, 523-549.
Fischbein, E. and Schnarch, D. (1997). The evolution with age of probabilistic, intuitively based
misconceptions. Journal for Research in Mathematics Education, 28(1), 96-105.
Forman, E. A. (1996). Learning mathematics as participation in classroom practice: Implications
of sociocultural theory for educational reform. In Steffe, L. P., Nesher, P., Cobb, P., Goldin,
G. A., and Greer, B. (Eds.), Theories of Mathematical Learning. Mahwah, NJ: Lawrence
Erlbaum Associates, Inc.
Franklin, C. A. and Garfield, J. B. (2006). The GAISE project: Developing statistics education
guidelines for grades pre-k-12 and college courses. In Gail F. Burrill and Portia C. Elliott
(Eds.), Thinking and Reasoning with Data and Chance: Sixty-eighth Yearbook (pp. 345375). Reston, VA: The National Council of Teachers of Mathematics, Inc.
Frederick University. (2008). Prospectus 2008-2009. Nicosia, Cyprus: International and Public
Relations Service, Frederick University.
Gal, I. (2004). Statistical literacy: Meanings, components, responsibilities. In D. Ben-Zvi and J.
Garfield (Eds.), The challenge of developing statistical literacy, reasoning and thinking (pp.
47-78). Netherlands: Kluwer Academic Publishers.
Garfield, J. (1993). Teaching statistics using small-group cooperative learning. Journal of
Statistics Education, 1(1).
Garfield, J. (1995). How students learn statistics. International Statistical Review, 63(1), 25-34.
Garfield, J. (2003). Assessing statistical reasoning. Statistics Education Research Journal, 2(1),
22-38.
317

Garfield, J. & Ahlgren, A. (1988). Difficulties in learning basic concepts in probability and
statistics: Implications for research. Journal of Research in Mathematics Education, 19(1),
44-63.
Garfield, J., del Mas, R., and Chance, B. (2006). Assessment resource tools for improving
statistical thinking – ARTIST online tests. Retrieved July 18, 2009 from
https://app.gen.umn.edu/artist/tests/index.html
Gigerenzer, G. (1991b). How to make cognitive illusions disappear: Beyond “heuristics and
biases.” In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol.
2, pp. 83-115). London: Wiley.
Giraud, G. (1997). Cooperative learning and statistics instruction. Journal of Statistics
Education, 5(3).
Green, D. R. (1979). The chance and probability concepts project. Teaching Statistics, 1(3), 6671.
Green, D. R. (1982a). Probability concepts in 11-16 year old pupils (2nd ed.). Centre for
Advancement of Mathematical Education in Technology, University of Technology,
Loughborough.
Green, D. R. (1982b). A survey of probability concepts in 3000 pupils aged 11-16 years. In D. R.
Grey, P. Holmes, V. Barnett, & G. M. Constable (Eds.), Proceedings of the First
International Conference on Teaching Statistics (pp. 766-783). Sheffield, England:
Teaching Statistics Trust.
Green, D. R. (1983). Shaking a six. Mathematics in School, 12, 5, 29-32.
Gunawardena, K. L. D. (1998). Introductory statistics: A cooperative learning approach. In
Proceedings of the International Conference on Teaching Statistics (ICOTS) 5, Singapore.
Retrieved
June
14,
2011
from
http://www.stat.auckland.ac.nz/~iase/publications/2/Topic2i.pdf
Hakes, R. R. (1998). Interactive-engagement versus traditional methods: A six-thousand-student
survey of mechanics test data for introductory physics courses. American Journal of
Physics, 66(1), 64-74.
Hall, M. R. and Rowell, G. H. (2008). Introductory statistics education and the national science
foundation. Journal of Statistics Education, 16(2).
Hawkes, N. (2006). Scientists find the key to cot deaths. The Times. Retrieved November 1,
2006, from http://www.timesonline.co.uk

318

Hiebert, J., and Carpenter, T. P. (1992). Learning and teaching with understanding. In Grouws,
D. A. (Ed.), Handbook of research on mathematics teaching and learning (pp. 65-97). New
York, USA: Macmillan.
Hiebert, J., and Lefevre, P. (1986). Conceptual and procedural knowledge in mathematics: An
introductory analysis. In Hiebert, J. (Ed.), Conceptual and procedural knowledge: The case
of mathematics, (pp. 1-27). Hillsdale, NJ: Lawrence Erlbaum Associates.
Hiebert, J., Stigler, J., Jacobs, J., Givvin, K., Garnier, H., Smith, M., et al. (2005). Mathematics
teaching in the United States today (and tomorrow): Results from the TIMSS 1999 video
study. Educational Evaluation and Policy Analysis, 27, 111-132.
Hirsch, L. S. and O’Donnell, A. M. (2001). Representativeness in statistical reasoning:
identifying and assessing misconceptions. Journal of Statistics Education, 9(2).
Ignatiou, K. and Zotos, E. (2007a). Common core mathematics for grade 12 of the unified
lyceum. Nicosia, Cyprus: Ministry of Education and Culture, Pedagogical Institute, Program
Development Services.
Ignatiou, K. and Zotos, E. (2007b). Mathematics of choice for grade 12 of the unified lyceum.
Nicosia, Cyprus: Ministry of Education and Culture, Pedagogical Institute, Program
Development Services.
Jacobs, V. R. (1999). How do students think about statistical sampling before instruction?
Mathematics Teaching in the Middle School, 5, 240-246.
Jaworski, B. (1998). The centrality of the researcher: Rigor in a constructivist inquiry into
mathematics teaching. In Teppo, A. R. (Ed.), Qualitative Research Methods in Mathematics
Education. Journal for Research in Mathematics Education Monographs, 9, 112-127.
Johnson, D., and Johnson, R. (n.d.). Cooperative learning. Retrieved June 5, 2009 from
http://www.co-operation.org/pages/cl.html
Johnson, D. W., and Johnson, R. T. (2004). Assessing students in groups: Promoting group
responsibility and individual accountability. Thousand Oaks, CA: Corwin.
Johnson, R. T., and Johnson, D. W. (1985). Student-student interaction: Ignored but powerful.
Journal of Teacher Education, 34(36), 22-26.
Johnson, D. W., Maruyama, G., Johnson, R., Nelson, D., and Skon, L. (1981). Effects of
cooperative, competitive, and individualistic goal structures on achievement: A metaanalysis. Psychological Bulletin, 89, 47-62.

319

Jolliffe, F. (2005). Assessing probabilistic thinking and reasoning. In Jones, G. A. (Ed.),
Exploring Probability in School: Challenges for Teaching and Learning (pp. 325-344). New
York, NY: Springer Science and Business Media, Inc.
Jones, D. L. (2004). Probability in middle grades mathematics textbooks: An examination of
historical trends, 1957-2004. Unpublished doctoral dissertation, University of Missouri,
Columbia.
Jones, P. S. (1970). A history of mathematics education in the United States and Canada.
Washington, DC: National Council of Teachers of Mathematics.
Jones, C. J. and Harris, P. L. (1982). Insight into the law of large numbers: A comparison of
Piagetian and judgment theory. Quarterly Journal of Experimental Psychology, 34A, 479488.
Jones, G. A., Langrall, C. W., and Mooney, E. S. (2007). Research in probability: Responding to
classroom realities. In Lester, F. K. Jr. (Ed.), Second Handbook of Research on Mathematics
Teaching and Learning: A Project of the National Council of Teachers of Mathematics.
Charlotte, NC: Information Age Publishing.
Jones, G. A., Langrall, C. W., Thornton, C. A., and Mogill, A. T. (1997). A framework for
assessing and nurturing young children’s thinking in probability. Educational Studies in
Mathematics, 32, 101-125.
Jones, G. A., Langrall, C. W., Thornton, C. A., & Mogill, A. T. (1999). Students’ probabilistic
thinking in instruction. Journal for Research in Mathematics Education, 30, 487-519.
Jones, G. A., and Thornton, C. (2005). An overview of research into the teaching and learning of
probability. In Jones, G. A. (Ed.), Exploring Probability in School: Challenges for Teaching
and Learning (pp. 65-94). New York, NY: Springer Science and Business Media, Inc.
Jones, G. A., Thornton, C. A., Langrall, C. W. and Tarr, J. E. (1999). Understanding students’
probabilistic reasoning. In Stiff, L. V. and Curcio, F. R. (Eds.), Developing Mathematical
Reasoning in Grades K-12: 1999 Yearbook. Reston, VA: The National Council of Teachers
of Mathematics, Inc.
Kahneman, D., Slovic, P., and Tversky, A. (1982). Judgment under uncertainty: Heuristics and
biases. Cambridge University Press.
Kahneman, D., and Tversky, A. (1972). Subjective
representativeness. Cognitive Psychology, 3, 430-454.

probability:

A

judgment

of

Kahveci, M. and Imamoglu, Y. (2007). Interactive learning in mathematics education: Review of
recent literature. Journal of Computers in Mathematics and Science Teaching, 26(2), 137153.
320

Kaplan, J. J. (2006). Factors in statistics learning: Developing a dispositional attribution model
to describe differences in the development of statistical proficiency. Doctoral dissertation,
The University of Texas at Austin.
Kaplan, M. and Kaplan, E. (2006). Chances are … Adventures in probability. NY: Penguin
Group, Inc.
Keeler, C. M. and Steinhorst, R. K. (1995). Using small groups to promote active learning in the
introductory statistics course: A report from the field. Journal of Statistics Education, 3(2).
Keeler, C. and Steinhorst, K. (2001). A new approach to learning probability in the first statistics
course. Journal of Statistics Education, 9(3). Retrieved April 24, 2009 from
http://www.amstat.org/publications/jse/v9n3/keeler.html
Khazanov, L. (2005). An investigation of approaches and strategies for resolving students’
misconceptions about probability in introductory college statistics. Doctoral dissertation,
Columbia University.
Khazanov, L. (2008). Addressing students’ misconceptions about probability during the first
years of college. Mathematics and Computer Education, 42(3), 180-192.
Knypstra, S. (2009). Teaching statistics in an activity encouraging format. Journal of Statistics
Education, 17(2).
Konold, C. (1989). Informal conceptions of probability. Cognition and Instruction, 6, 59-98.
Konold, C. (1993). Inconsistencies in students’ reasoning about probability. Journal for
Research in Mathematics Education, 24(5), 392-414.
Konold, C. (1995). Issues in assessing conceptual understanding in probability and statistics.
Journal of Statistics Education, 3(1).
Konold, C., Pollatsek, A., Well, A., Lohmeier, J., and Lipson, A. (1993). Inconsistencies in
students’ reasoning about probability. Journal for Research in mathematics Education,
24(5), 392-414.
Krause, E. C. (2001). The effect of instruction in representation theory on student understanding
of binomial distribution. Doctoral dissertation, University at Albany, State University of
New York.
Kvatinsky, T., and Even R. (2002). Framework for teacher knowledge and understanding about
probability. In Phillips, B. (Ed.), Proceedings of the Sixth International Conference on
Teaching Statistics (ICOTS 6).
321

Langrall, C. W. and Mooney, E. S. (2005). Characteristics of elementary school students’
probabilistic thinking. In Jones, G. A. (Ed.), Exploring Probability in School: Challenges for
Teaching and Learning (pp. 95-120). New York, NY: Springer Science and Business Media,
Inc.
Laplace, P. S. (1995). Theorie analytique des probabilities [Analytical theory of probabilities].
Paris: Jacques Gabay. (Original work published 1814).
Lecoutre, M. P. (1992). Cognitive models and problem spaces in “purely random” situations.
Educational Studies in Mathematics, 23, 557-568.
Leikin, R. and Zaslavsky, O. (1997). Facilitating student interactions in mathematics in a
cooperative learning setting. Journal for Research in Mathematics Education, 28(3), 331354.
Levine, D. M., Krehbiel, T. C., and Berenson, M. L. (2010). Business statistics: A first course
(5th ed.). Upper Saddle River, NJ: Pearson Education, Inc.
Li, J. (2000). Chinese students’ understanding of probability. Doctoral dissertation, Nanyang
Technological University.
Liu, Y. and Thompson, P. (2007). Teachers’ understanding of probability. Cognition and
Instruction, 25(2), 113-160.
Lordou-Kaspari, D. (2003). Curriculum development in mathematics in Cyprus secondary
schools (1960-1995). Nicosia, Cyprus: Livadiotis Ltd.
Mackisack, M. (1994). What is the use of experiments conducted by statistics students? Journal
of Statistics Education, 2(1).
Manage, A. B. W. and Scariano, S. M. (2010). A classroom note on: Student misconceptions
regarding probabilistic independence vs. mutual exclusivity. Mathematics and Computer
Education, 44(1), 14-21.
Mathematical Association of America. (1998). Quantitative reasoning for college graduates: A
complement to the standards. Retrieved May 10, 2009 from http://www.maa.org/past/
ql/ql_toc.html
McConnell, J. W., Brown, S., Usiskin, Z., Senk, S. L., Widerski, T., Anderson, S., Eddins, S.,
Feldman, C. H., Flanders, J., Hackworth, M., Hirschhorn, D., Polonsky, L., Sachs, L., and
Woodward, E. (1998). The University of Chicago School Mathematics Project: Algebra:
Integrated Mathematics. Glenview, IL: Scott, Foresman and Company.
Ministry of Education and Culture, Cyprus. (2004). Country report: Cyprus. Retrieved April 26,
2009 from http://www.coe.int/T/DG4/Linguistic/Source/Country_Report_ Cyprus_ EN.pdf
322

Ministry of Education and Culture & Pedagogical Institute. (2002). Post-secondary education in
Cyprus. Nicosia, Cyprus: Program Development Service.
Ministry of Education and Culture, Pedagogical Institute & Curriculum Development Service
(2010). Program of Studies: Mathematics. Retrieved March 12, 2011 from
http://www.moec.gov.cy/analytika_programmata/
Mises, R. von (1952). Probability, statistics, and truth (J. Neyman, O. Scholl, & E. Rabinovitch,
Trans.). London: William Hodge and company. (Original work published 1928).
Mojica, G. F. (2006). Middle school teachers’ conceptions of empirical and theoretical
probability: A study of pedagogy. Dissertation, North Carolina State University, NC.
Montgomery, D. (1997). Design and analysis of experiments (4th ed.). New York, NY: John
Wiley and Sons, Inc.
Moore, D. S. (1997). Probability and statistics in the core curriculum. In J. Dossey (Ed.),
Confronting the Core Curriculum (pp. 93-98). Mathematical Association of America.
Mullis, I. V. S., Martin, M. O., and Foy, P. (2008). International mathematics report: Findings
from IEA’s trends in international mathematics and science study at the fourth and eighth
grades. Chestnut Hill, MA: TIMSS & PIRLS International Study Center, Lynch School of
Education, Boston College.
National Commission on Excellence in Education. (1983). A nation at risk: The imperative for
educational reform. Retrieved October 25, 2006, from http://www.ed.gov/pubs/Nat
AtRisk/recomm.html
National Council of Supervisors of Mathematics. (1977). Position paper on basic skills.
Arithmetic Teacher, 25(1), 19-22.
National Council of Teachers of Mathematics. (1989). Curriculum and evaluation standards for
school mathematics. Reston, VA: Author.
National Council of Teachers of Mathematics. (1991). Professional standards for teaching
mathematics. Reston, VA: Author.
National Council of Teachers of Mathematics. (2000). Principles and standards for school
mathematics. Reston, VA: Author.
National Research Council. (2003). Evaluating and improving undergraduate teaching in
science, technology, engineering and mathematics. Washington, DC: National Academies
Press.
323

National Research Council. (2001). Adding + it up. Kilpatrick, J., Swafford, J., and Findell, B.
(Eds.). Washington, D.C.: National Academy Press.
Neapolis University. (2011). Neapolis University Pafos. Retrieved January 12, 2012
http://www.nup.ac.cy/Portals/0/Publications/NUP_Brochure_EN_lowRes.pdf
Nickerson, R. S. (2004). Cognition and chance: The psychology of probabilistic reasoning.
Mahwah, NJ: Lawrence Erlbaum Associates, Publishers.
Open University of Cyprus.
(2011).
http://www.ouc.ac.cy/web/guest/home

Retrieved

October

15,

2011

from

Papanastasiou, C. (1997). Cyprus. In D. Robitaille (Ed.), National contexts for mathematics and
science education: An encyclopedia of the education systems participating in TIMSS (p. 9197). Canada: Pacific Educational Press.
Papanastasiou, C. (2002). TIMSS study in Cyprus: Patterns of achievements in mathematics and
science. Studies in Educational Evaluation, 28(2002), 223-233.
Papapavlou, A. N. (2001). Linguistic imperialism? The status of English in Cyprus. Language
Problems & Language Planning, 25(2), 167-176.
Pashiardis, P. (2007). Cyprus. In Horner, W., Dobert, H., Kopp, B. von, & Mitter, W. (Eds.), The
Education Systems of Europe (pp. 202-222). Downloaded September 12, 2009 from
http://www.springerlink.com/content/hq08um126n636423/
Penas, L. M. (1987). Probability and statistics in midwest high schools. In American Statistical
Association 1987 Proceedings of the Section on Statistical Education, (pp. 122). Alexandria,
VA: American Statistical Association.
Pfaff, T. J. and Weinberg, A. (2009). Do hands-on activities increase student understanding? A
case study. Journal of Statistics Education, 17(3).
Piaget, J., and Inhelder, B. (1975). The origin of the idea of chance in children (L. Leake Jr., P.
Burrell, & H. D. Fischbein, Trans.). New York: Norton (Original work published 1951).
Polaki, M. V., Lefoka, P. J., and Jones, G. A. (2000). Developing a cognitive framework for
describing and predicting Basotho students’ probabilistic thinking. Boleswa Educational
Research Journal, 17, 1-21.
Pollatsek, A., Well, A. D., Konold, C., & Hardiman, P. (1987). Understanding conditional
probabilities. Organizational Behavior and Human Decision Processes, 40, 255-269.
Potthast, M. J. (1999). Outcomes of using small-group cooperative learning experiences in
introductory statistics courses. College Student Journal, 33(1).
324

Pratt, D. (2000). Making sense of the total of two dice. Journal for Research in Mathematics
Education, 31(5), 602-625.
Resnick, L. B., & Ford, W. W. (1981). The psychology of mathematics for instruction. Hillsdale,
NJ: Erlbaum
Rubel, L. H. (2006). Students’ probabilistic thinking revealed: The case of coin tosses. In G. F.
Burrill & P. C. Elliott (Eds.), Thinking and Reasoning with Data and Chance (pp. 49-59).
Reston, VA: National Council of Teachers of Mathematics.
Rubel, L. H. (2007). Middle school and high school students’ reasoning on coin tasks. Journal
for Research in Mathematics Education, 38(5), 531-556
Saenz, C. (1998). Teaching probability for conceptual change. Educational Studies in
Mathematics, 35, 233-254.
Savery, J. R. and Duffy, T. M. (1996). Problem based learning: An instructional model and its
constructivist framework. In B. G. Wilson (Ed.), Constructivist Learning Environments:
Case Studies in Instructional Design. Englewood Cliffs, NJ: Educational Technology
Publications.
Scheaffer, R. L., Watkins, A. E., & Landwehr, J. M. (1998). What every high-school graduate
should know about statistics. In S. P. Lajoie (Ed.), Reflections on statistics: Learning,
teaching, and assessment in grades K-12. Mahwah, NJ: Lawrence Erlbaum Associates.
Sedlmeier, P. and Gigerenzer, G. (1997). Intuitions about sample size: The empirical law of large
numbers. Journal of Behavioral Decision Making, 10(1), 33-51.
Shaughnessy, J. M. (1977). Misconceptions of probability: An experiment with a small-group,
activity-based model building approach to introductory probability at the college level.
Educational Studies in Mathematics, 8(3), 295-316.
Shaughnessy, M. (1981). Misconceptions of probability: From systematic errors to systematic
experiments and decisions. In A. P. Shulte and J. R. Smart (Eds.), Teaching statistics and
probability (pp. 90-99). Reston, VA: National Council of Teachers of Mathematics.
Shaughnessy, M. (1992). Research in probability and statistics: Reflections and directions. In D.
A. Grouws (Ed.), Handbook of research on mathematics teaching and learning (pp. 465494). Reston, VA: National Council of Teachers of Mathematics.
Shaughnessy, M. (2003). Research on students’ understanding of probability. In J. Kilpatrick, W.
G. Martin, D. Schifter (Eds.), A research companion to principles and standards for school
mathematics (pp. 216-226). Reston, VA: National Council of Teachers of Mathematics.
325

Shaughnessy, M. J., Barrett, G., Billstein, R., Kranendonk, H. A., and Peck, R. (2004). What’s
the probability of a hit? In Lott, J. W. and House, P. A. (Eds.), Navigating through
probability in grades 9-12, p. 40-42 and p. 94-97. Reston, VA: The National Council of
Teachers of Mathematics, Inc.
Shay, K. B. (2008). Tracing middle school students’ understanding of probability: A longitudinal
study. Doctoral dissertation, Rutgers, The State University of New Jersey.
Silver, E. (1986). Using conceptual and procedural knowledge: A focus on relationship. In
Hiebert, J. (Ed.), Conceptual and procedural knowledge: The case of mathematics (pp. 181198). Hillsdale, NJ: Lawrence Erlbaum Associates.
Silver, E. A., Mesa, V. M., Morris, K. A., Star, J. R., and Benken, B. M. (2009). Teaching
mathematics for understanding: An analysis of lessons submitted by teachers seeking
NBPTS certification. American Educational Research Journal, 46(2), 501-531.
Skemp, R. (1978a). The psychology of learning mathematics. Hillsdale, NJ: Lawrence, Erlbaum.
Skemp, R. R. (1978b). Relational understanding and instrumental understanding. Arithmetic
Teacher, 26(3), 9-15.
Slavin, R. E. (1995). Cooperative learning: Theory, research, and practice (2nd ed.). Boston:
Allyn & Bacon.
Smith, K. A., Douglas, T. C. and Cox, M. F. (2009). Supportive teaching and learning strategies
in STEM education. New Directions for Teaching and Learning, 117, 19-32.
Springer, L., Stanne, M. E., and Donovan, S. S. (1999). Effects of small-group learning on
undergraduates in science, mathematics, engineering, and technology: A meta-analysis.
Review of Educational Research, 69(1), 21-51.
Stanovich, K. E. (1999). Who is rational? Studies of individual differences in reasoning.
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Stanovich, K. E., and West, R. F. (2003). Evolutionary versus instrumental goals: How
evolutionary psychology misconceives human rationality. In D. E. Over (Ed.), Evolution
and the psychology of thinking: The debate. Retrieved May 10, 2009 from
http://web.mac.com/kstanovich/iWeb/Site/Research%20on%20Reasoning.html
Star, J. R. (2005). Reconceptualizing procedural knowledge. Journal for Research in
Mathematics Education, 36(5), 404-411.
Tarr, J. E., and Jones, G. A. (1997). A framework for assessing middle school students’ thinking
in conditional probability and independence. Mathematics Education Research Journal, 9,
39-59.
326

The Ministry of Education and Culture, Republic of Cyprus. (2008). Inclusion in the Cyprus
educational system at the beginning of the twenty first century: An overview - national
report of Cyprus. Nicosia, Cyprus: Author.
Thompson, P. W., and Saldanha, L. A. (2003). Fractions and multiplicative reasoning. In J.
Kilpatrick, G. Martin, & D. Schifter (Eds.), A research companion to the principles and
standards for school mathematics. Reston, VA: National Council of Teachers of
Mathematics.
TIMSS International Study Center. (2000). Mathematics teacher background with mathematics
achievement: Main survey, Third International Mathematics and Science Study Repeat.
Retrieved July 4, 2009, from http://timss.bc.edu/timss1999i/questionnaires.html
TIMSS and PIRLS International Study Center. (1995). TIMSS: IEA’s third international
mathematics and science study: Released item set for the final year of secondary school:
Mathematics and science literacy, advanced mathematics, and physics. Retrieved July 4,
2009, from http://timss.bc.edu/timss1995i/Items.html
TIMSS and PIRLS International Study Center. (2001). TIMSS 1999 mathematics items: Released
set for eighth grade. Retrieved July 4, 2009 from http://timss.bc.edu/timss 1999i/study.html
TIMSS and PIRLS International Study Center (2007). TIMSS 2003 mathematics items: Released
set eighth grade. Retrieved July 6, 2009 from http://timss.bc.edu/timss2003i/ released.html
TIMSS and PIRLS International Study Center (2009). TIMSS 2007 user guide for the
international database. Retrieved July 6, 2009 from http://timss.bc.edu/TIMSS2007/
items.html
Tinto, V. (1993). Leaving college: Rethinking the causes and cures of student attrition (2nd ed.).
Chicago: University of Chicago Press.
Travers, K. J., & Westbury, I. (1989). The IEA study of mathematics I: Analysis of mathematics
curricula. New York: Pergamon Press.
Tversky, A. and Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.
Science, 185, 1124-1131.
Ulep, S. A. (1990). Strategies preservice secondary mathematics teachers use in solving
problems involving uncertainty. Dissertation Abstracts International, (UMI No. 9116999).
University of Central Lancashire-Cyprus. (2012). UCLAN Cyprus: The first British university in
Cyprus. Retrieved May 1, 2012 from www.uclancyprus.ac.cy

327

University of Cyprus. (2007). Guide to undergraduate studies 2007-2008. Retrieved May 18,
2009 from http://www.ucy.ac.cy/data/puof/undgradgr08.pdf
University of Nicosia. (2008). Profile 2008-2009. Nicosia, Cyprus: Author.
Utts, J. (2003). What educated citizens should know about statistics and probability. The
American Statistician, 57(2), 74-79.
Vrasidas, C. and McIsaac, M. (2001). Integrating technology in teaching and teacher education:
Implications for policy and curriculum reform. Educational Media International, 38(2), 127132.
Watson, J. (2005). The probabilistic reasoning of middle school students. In Jones, G. A. (Ed.),
Exploring Probability in School: Challenges for Teaching and Learning (pp. 145-170). New
York, NY: Springer Science and Business Media, Inc.
Watson, J. M. (2006). Statistical literacy at school: Growth and goals. NJ: Lawrence Erlbaum
Associates Inc.
Watson, J. M., & Moritz, J. B. (2002). School students’ reasoning about conjunction and
conditional events. International Journal of Mathematics Education in Science and
Technology, 33(1), 59-84.
Webb, N. M. (1991). Task-related verbal interaction and mathematics learning in small groups.
Journal for Research in Mathematics Education, 22(5), 366-389.
Webb, N. M., Tropper, J. D., and Fall, R. (1995). Constructive activity and learning in
collaborative small groups. Journal of Educational Psychology, 87(3), 406-423.
Woolfson, M. M. (2008). Everyday probability and statistics: Health, elections, gambling and
war. London, UK: Imperial College Press.
Zieffler, A., Garfield, J., Alt, S., Dupuis, D., Holleque, K., and Chang, B. (2008). What does
research suggest about the teaching and learning of introductory statistics at the college
level? A review of the literature. Journal of Statistics Education, 16(2).
Zimmermann, G. (2002). Students’ reasoning about probability simulations during instruction.
Dissertation, Illinois State University.

328