AN ANALYSIS OF THE EDUCATIONAL TESTING
PROGRAM OF THE SCHOOL OF POLICE
ADMINISTRATDN AND PUBUC SAFETY

Thesis for the Degree of Ma S.
MICHlGAN STATE UNIVERSWY

Donaid Jameson Pope
1966

 

, Y
R
A
R
B
I
L

m
.n
m
is
man
M

.w.
m
V
.m
U

 

III/7T '

w

W W

I

[9
2

1

m3

W

H

mm

AN ANALYSIS OF THE EDUCATIONAL TESTING PROGRAM
OF THE SCHOOL OF POLICE ADMINISTRATION

AND PUBLIC SAFETY

By

Donald Jameson Pope

AN ABSTRACT OF A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Ibpartment of Police Awninistration and Pablic Safety

May 1966
APPROVEDJ rm / WWW

haifman)

       
 

kw

éTember)

ABSTRACT

AN ANALYSIS OF THE EDUCATIONAL TESTING PROGRAM
OF THE SCHOOL OF POLICE ADMINISTRATION
AND PUBLIC SAFETY

by Donald Jameson Pepe

For a total of eleven years the School of Police Administra-
tion and Public Safety at Michigan State University administered a
battery of tests to all freshmen and transfer students entering
the school. This test battery was composed of tests designed to
measure academic ability, emotional maturity and stability, and
vocational interest. The purpose of this study has been to evalu-
ate the effectiveness of this testing program. This evaluation
'was approached from two angles. The first of these was the over—
all effectiveness of the program and the use made of its results.
The second approach concerned the actual data gathered, and how
this data might be used to predict the academic success of stu-
dents enrolling in the School of Police Administration and Public
Safety. As the reader will see in succeeding chapters, the study
presents some conclusions, but on other points the study has been
inconclusive or contradictory.

An evaluation of the whole testing program was not possible
because of the somewhat deficient planning and organization behind

the program. Those conclusions that can be drawn indicate fairly

Donald Jameson Pepe

clearly that the program was only a qualified success. There was
little use made of the data as it was gathered, and it was not organ-
ized in a manner to facilitate any study of its over-all meaning or

L contribution.

Since the total program could not be studied, conclusions as
to the predictive ability of its results must be based on a rela-
tively small portion of the data gathered. These conclusions are
somewhat tentative and do little more than show that one of the
tests studied, the College Qualification Test, bears a small rela-
tionship to academic success. This relationship is consistent, but
not strong enough to justify its use as a device for screening
applicants for admission.

waever, the shortcomings of this first testing program do
not mean that further efforts should not be considered. If any-
thing, it points out our lack of knowledge in areas where we should
know much more than we presently do know. There is a very real need
for more exhaustive studies in areas relative to success in higher
education. The School of Police Administration and Public Safety
can and should assume leadership in studying those areas related to

its own unique contribution to the community.

AN ANALYSIS OF THE EDUCATIONAL TESTING-PROGRAM
OF THE SCHOOL OF POLICE ADMINISTRATION
AND PUBLIC SAFETY

By

Donald Jameson Pepe

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Police Administration and Public Safety

May 1966

ACKNOWLEDCMENTS

If I were to give full and detailed credit to all who con-
tributed to the development of this thesis, the acknowledgments
would be longer than the text. This campus seems to have more than
its quota of helpful people who courteously answer what very prob-
ably seem foolish questions to them.

The staff at the computer center was most helpful, especi—
ally, Mrs. Nbrma Landis. I am deeply indebted to Doctor Karslake
of the Psychology Department for his counseling on matters method-
ological. This study could not have been completed without his
invaluable assistance. my special appreciation to Doctor Leon
Wbaver who has assisted me throughout this study, listened to my
tales of woe, and encouraged me to further efforts; and finally,
deepest gratitude to my mother, who is even now typing the last

pages of text as I write these acknowledgements.

ii

Chapter

I.

II.

III.

IV.

TABLE OF CONTENTS

INTRODUCTION . . . .

Contributions of the Study

Limitations of the SRudy

Methodology . . .

Organization of the Remainder

Thesis . . .

THE TESTING PROGRAMC .

O

The M.S.U. Orientation Tbsts

Otis That of Mental Ability

The Minnesota Multiphasic Personality

Inventory . . .

A.REVIEW OF CURRENT ATTEMPTS TO PREDICT

ACADEMIC PERFORMANCE

Specialized Test Batteries .

High School Performance as an

Multiple Factor Correlations

General Comments .
STUDY DESIGN . . . .

Preliminary Problems

The Final Sample .

Criteria of Academic
iii

Success

9

Indication

I-4
II-1
II-2

II-4

II~5

III-1
III-9
III—13
III-13
III-15 _
IVF1
IV¥1
IV43

IV-S

Chapter

Recording Procedure . . . . . . .
Study Design . . . . . . . . .
Implementing the Design . . . . .

Computer Program . . . . . . . .

V. RESULTS OF STUDY . . . . . . . . .

The First Computer Run . . . . . .
The Second Computer Run . . . . .
Cross validation . . . . . . . .

Cross Checking for Bias . . . . .

VI. ANALYSIS OF RESULTS AND CONCLUSIONS . . .

BIBLIOGRAPHY
APPENDIX .

Comparison with Other Studies . . .
The meaning of the Statistical Results.
Evaluation of the Tbsting Program . .

Suggestions for Future Shudies . . .

iv

LIST OF TABLES

Table Page

1. Correlation of Selected Variables with the GPA's

of Freshmen in a Liberal Arts College . . . . . III-4
2. Comparison of Correlations Based on Ability and

Attitudes with First Term and First Full Year

College Success . . . . . . . . . . . . III-5
3. Correlation of Certain Tests and sub-Test Scores

with the First Term Grades of Police Adminis—

tration Freshmen . . . . . . . . . . . . III-7
4. Correlations of Freshmen GPAfs with TWO Scales of

the Scholastic Aptitude Tbst . . . . . . . . III-8
S. GPA Correlations with a Specially Designed Test

for Predicting Success in a College of Law . . . III—10
6. Test Correlations with Grades in Engineering and

Non-Engineering Courses for Engineering Majors . . III-11
7. Differential Prediction in Three Types of

College Curricula . . . . . . . . . . . III—12
8. High School Rank and GPA Compared with Other

PrediCtor variables 0 e e o e e o e e e e III—14

Table Page
10. Number of Subjects with Complete Scores in

EachTerm.............. IV-4
110 Division of Term sub-Groups into the Three

Computational Steps . . . . . . . . . . IV—9
12. Ayerages of variables and Correlations with All—

University and Police Administration Grade

Point Averages . . . . . . . . . . . . V41
13. Correlations of Selected Variables with the Two

Grade Point Averages and with Each Other -

First Computer Run . . . . . . . . . . . V—5
14. A Comparison of Selected Portions of the Results

of the First and Second Computer Runs . . . . V—7
15. Correlations of Selected Variables with the TWO

Grade Point Averages and with Each Other —

Second Computer Run . . . . . o . . . . Vm8
16. Cross Validation of Regression Equations: R's

between Predicted and Achieved GPA's and

Z Values of Changes . . . . . . . . . . V411
17. Simple Correlations between CQT Total and the TWO

GPA's and Results of Cross Validation of the

RegreSSion Equation 0 e e e o o o o o e V_13

vi

CHAPTER ONE

INTRODUCTION

For a period of eleven years, ending with the Fall term of
1963, the School of Police Administration and Public Safety at Michi-
gan State University administered a mental and psychological testing
program to all entering students. The purpose of this study is to
evaluate the effectiveness and significance of that testing program.
The most important part of this study centers around the search for
components of the test battery that will predict academic success.
There will also be occasion to evaluate the testing program in terms
of how well it was organized and the uses to which its results were
put.

Nb formal hypothesis is tested in this study. Rather, this
study represents an attempt to organize an already existing body of
data and to search for possible relationships between the data and

academic success in college.

CONTRIBUTIONS OF THE STUDY

The importance of this study lies in its contributions to two
major areas. The first of these is in providing the School of
Police Administration and Public Safety with an evaluation of their
new discontinued testing program. It is hoped that, on the basis
of this evaluation, the school will be encouraged to develop better

organized and more effective testing programs in the future. The

I-2

second contribution, hopefully, will be to our general knowledge
and understanding of the factors relevant to academic success.

At the present time, the search for factors relating to aca-
demic success and attempting to predict academic success receive
high priorities from psychologists and educators. Their concern
for this area is not without justification, as we know all too
little about why some students succeed and others fail in school.
we know even less about why some seemingly good students fail while
many poorer students succeed. To be sure, we can identify many of
the relevant factors operating here, but we are not yet able to

determine how and to what extent these factors operate.

LIMITATIONS OF THE STUDY

Any conclusions drawn on the basis of this study will neces-
sarily be limited by the design of the testing program and by the
students participating in the program. The design of the testing
program imposes several limitations on the study. The first of
these is that the complete program could not be studied. One
reason for this is that the composition of the test battery was
changed in the middle of the eleven-year span of the program, thus
making difficult or impossible an evaluation of test results cover-
ing the entire eleven year period. The other reason is that many
of the students' test records were imcomplete, and therefore un-
usable.

Another possible limitation lies in the fact that the stu-
dents taking the tests are not representative of the University
population as a whole. In fact, the Office of Evaluation Services

reports on comparative standings on the M. S. U. Orientation Tests

I-3

show that students entering the School of Police Administration and
Public Safety have average scores much below the University average.
The only academic areas scoring consistently lower than Police
Administration were Agriculture and Education.

As a result of these limitations, any efforts to extend the
conclusions of this study to other testing programs or to other stu-
dent groups should proceed with caution. There will be some gener-
alisations made in the concluding sections of this study, but they
will be somewhat tentative and in general supported by the findings

of other researchers.

METHODOLOGY

The general plan of the study was first to divide the records
from the testing program into three equal groups, with each group
containing an equal number of students from each term. Then,
through correlating test scores with academic performance, as repre-
sented by all-University and Police Administration Grade Point
Ayerages, the first two groups of students were used to identify
test variables which seemed to correlate well with performance.
These remaining test variables were then used to predict academic
performance for the third group of students. A.comparison of
their predicted performance with their actual performance would
indicate, hepefully, the extent to which the test items related to
and.were indicative of, a student's level of academic success.

The reader will note in succeeding chapters that the neat
study plan outlined above did not always prove workable. However,
'with modifications in design, the study was still able to reach

some meaningful conclusions.

1-4

ORGANIZATION OF THE REMAINDER OF THIS THESIS

The remainder of this thesis is organized into five major
parts, each part represented by a separate chapter. The next
chapter (Chapter II) deals with the nature of the Police Adminis-
tration testing program. Chapter III is based on a review of the
literature concerning attempts to predict academic success on the
college level. Chapter IV'provides a detailed explanation of the
study design. It also covers most of those factors that necessi-
tated changes in the design of the study.

Chapters V and VI deal with the actual results of the study,
the former representing the results and the latter analyzing them.
In Chapter VI conclusions are stated insofar as generalizations
from the data can be justified. The final chapter, in addition to
evaluating the testing program, also contains some suggestions con-
cerning content and organization of future studies that the School

of Police Administration and Public Safety might attempt.

II-1

CHAPTER TWO
THE TESTING PROGRAM

Beginning with the fall term of 1952 and continuing through
the fall term of 1963, all new freshmen and transfer students enter-
ing the School of Police Administration and Public Safety were given
a battery of psychological tests and inventories. The tests were
generally given in the introductory course in Police Administration
and then scored and interpreted by personnel in the Michigan State
University Counseling Center. The final results for each student,
reduced to summary form, were then transmitted back to personnel in
the School of Police Administration concerned with academic advis-
ing. These test scores were intended to aid in the advising pro-
cess by providing indications as to where each student's strong and
weak points lay. (The Appendix shows the report form employed).

In the course of these eleven years, about 2,200 students
were enrolled in the introductory Police Administration course. For
reasons that have not been documented, almost half of these students
had incomplete or totally missing test scores. There are about
1,200 partial or complete sets of test scores available, but as
tests taken over some periods were unusable, no exact count of the
total was made.

During this eleven year period there were a number of changes

made in both the composition of the test battery and in the form of

II-2

reporting. It is for this reason that the sample used in this
study is so much smaller than the number of students taking the
tests. Chapter four explains in detail why the nature of the
study necessitated selecting such a relatively small group.

The tests employed in this program were concerned'with
assessing emotional stability, maturity, vocational preference,
level of academic potential or development, and certain aspects of
the individual's personality. 3y covering such a wide scope, the
tests employed in this program represent a fair cross section of
the testing instruments available. Chapter' four lists all the
tests and inventories employed throughout the duration of the pro-
gram. A.brief review of the nature of these tests actually covered
in this study will provide the reader with some idea of its scope

and composition.

TE! 1!. S. U. ORIENTATION TESTS

The I; S. U. Orientation tests comprise a battery of tests
presented by the university to all incoming Freshmen and transfer
students. In both form and function they are similar to the entrance
tests employed by most colleges and universities. They are all
attempts to measure the general academic development of the entering
college student. During the test period the Uhivesity employed two
different batteries of entrance tests, but this study was limited
to just one of them.

1
The "College Qualification Tests” consist of a three-part

 

 

1
Cells alification Tbsts: Combined Booklet Idition, the
Psychologican Corporation, it. ThrE, l§5§.

II—3

multiple choice examination covering verbal ability, mathematical
ability and general information. In measuring verbal ability, the
student is asked to pick a word.that means the same (or opposite)
of a given word. Selection 1. made from a list of four alternatives
and the items vary in difficulty. In the mathematics section the
student is given a problem and asked to select the solution from a
list of alternatives. The general information section simply asks
questions drawn from diverse areas and asks the student to choose
the correct answers. The sum of these three tests, the CQT total
score, is also reported.

In addition to the COT the entrance battery also includes the
TISU Reading That"2 and the ”English Placement Test”? The reading
test is designed to measure a student's ability to understand and
learn from what he reads. It consists of eight reading selections
representative of several academic areas here at lichigan State
University. After each reading passage there is a series of ques-
tions on the meaning and contents of the passage. In addition to
measuring the student's ability to remember what he reads, the
questions also provide some measure of the factors involved in
critical thought.

The English Placement Test consists of thirtybfive multiple
choice questions representing various aspects of English usage:

spelling, capitalisation, grammar, punctuation, sentence structure

 

zssu Readi That, Office of Evaluation Services, Iichigan
State versity, t Lansing, Hichigan, Form 1962.

3 lish Placement Test, Benjamin B. Hickok, Department of
herican Thought and W, Hichigan State University, East
Lansing, 1962.

11-4

and Organisation. This test is used primarily to identify students
‘who need remedial assistance in English or who qualify for honors

sections.

OTIS TEST OF lllTlL ABILITT4

The Otis Test of lental Ability is designed to measure reason-
ing ability, not accumulated learning or knowledge. Ihile it is not
possible to measure mental ability directly, one can measure the
effect mental ability has on a student's ability to acquire certain
knowledge and.mental skill. One way of doing this is to ask ques-
tions that draw more on a person's innate ability and reasoning
power, and less on prior learning. It is admittedly impossible to
separate completely inborn skills from those acquired through learn-
ing processes, but a careful selection of test items can reduce this
problem to a point where meaningful statements can be made about a
person's mental ability.

This mental ability is scored and interpreted in terms of an
intelligence quotient, or IQ. The IQ,score is a function of the
ratio of mental age divided by chronological age, multiplied by 100
for ease of numerical representation. Hence a person with a mental
age of eighteen and a chronological age of eighteen would have an IQ
of 100 which is regarded as ”normal”. Shailarly, a.person'with a
mental age higher than his chronological age would have an above
average IQ (over 100).

This particular IQ test consists of eighty multiple choice

 

4Otis, Arthur 8., Otis ick-Scori lbntal‘nbilit Toots:
low Edition: Gamma That:
ork , l 54.

  

II—S

items dealing with vocabulary, arithmetic reasoning, spatial re-

lations, number series, logical relationships and other areas.

on nmsou mnpmsxc PERSONALITY Immomr (1mm)5

The IIPI was designed to provide a broad analysis of an indi-
vidual's personality structure in a single test instrument. As the
test was develOped it was administered to persons in mental institup
tions and to others with known mental or emotional problems. The
various scales on the test were formulated on the basis of patterns
of answers that correlated well with known and diagnosable mental
problems. The instrument consists of 550 statements, all in the
first person, covering a range from the physical condition of the
person being tested to his moral and social attitudes. Bach ques-
tion is to be answered Tyes”, ”no" or ”can't say". After the sub-
ject has answered all the questions about himself, his test is
scored so as‘to routinely yield scores on three or four validity
scales and ten or more clinical scales. The number of scales
scored will depend on the reason for giving the test. For the
School of Police Administration and Public Safety scoring was based
on three validity scales and nine clinical scales.

The L Score is based on a small group of items that will yield
a score tending to indicate any naive attempt by the subject to put
himself in a good light. High scores are obtained by persons who try
(often unconsciously) to answer all questions in ways that will fit

their conception of the moral code regardless of their own.know1edge

 

5Hathaway, Starks R. , and lchinley, Charnley J. , Booklet for

the linnesota lulti hasic Personalit Invente . (The PsychoIogical
Corporation, lbw Tbrk’, l§55.

11-6

of themselves. The questions usually dealt with undesirable con-
duct or moral “failures” that may seem bad, but which are common in
virtually all people to one degree or another. The assumption is
that a normal person, answering truthfully, will admit yes to most
of these items. Attempts at concealment here are called defensive-
ness, or even ”faking good” in extreme cases.

The P Score is just the opposite of the ”L" score. Persons
scoring high on this scale are often attempting'to show themselves
in a bad light. This could be because the subject is overly candid
or because he is generally maladjusted.

The I score is similar to the "L" score, but much.more subtle.
It is actually mixed in with five of the clinical scales which when
high indicate defensiveness and a lack of candor. Low scores, like
a high "F" score, indicate frankness and self criticism.

The Es egal=.is based on a comparison of normal persons with
patients having numerous physical complaints, many of them psycho-
logical in origin. These complaints may be organically based, but
are more often classed as psychosomatic.

The D scale is based on the scores of emotionally depressed
persons. A.high score here is often associated with persons who are
unsure of themselves and of the future. They frequently tend to be
sad or moody. Higher scores here frequently occur when a subject is
in trouble.

The E: scale is similar to the Rs scale, but gives evidence of
more elaborate and serious symptoms. High scores here are indica-
tive of persons who are more likely to avoid or run from crises and

when pushed too far are likely to develop physical or psychological

Ila?

illness as the means of escape from a difficult situation.

The Pd scale deals with what is termed psychopathic deviancy.
This is characterized by a person's failure to be controlled by the
mores of society and an apparent inability to feel guilt or remorse
or to respond to punishment. These pecple frequently commit asocial
acts with no apparent motive.

The Mf scale is a measure of masculine or feminine interests.
High scores here tend to indicate that a person's interests are more
in line with those of the Opposite sex than those of his or her own
sex. High scores here can also be indicative of homosexual tendenp
cies, especially for men.

The Pa scale is a measure of interpersonal sensitivity, and in
extreme cases, paranoia. High scores here may indicate a person who
feels he is being persecuted or mistreated by other pecple.

The Pt scale is related to compulsions and obsessions. High
scores here may indicate neurotic compulsions in selected areas,
excessive worrying or a tendency to be overly introspective and self
critical. many asocial acts are not a result of psychopathic devim
ancy, but of persons driven by compulsions or obsessions.

The Sc scale is concerned with how much the subject reacts and
and thinks like others around him. Extreme scores may be indicative
of IChIZOphrOnin, but more generally the scale measures a person's
tendency to distort what he sees and hears and to react in unusual
‘ways.

The Ha scale attempts to measure enthusiasm and energy. High
scorers become readily interested in new projects and approach prob-

lens with animation. In extreme cases this may result in antisocial

II-B

acts or irrational manic behavior.

It should be emphasized that the MMPI is not a complete diagb
nostic tool, but rather, that it will tend to indicate where a per—
son.may have emotional problems. Further testing and interviewing
is usually necessary before a diagnosis can be made.

This testing program was comprehensive in scape. As was
already mentioned, it dealt not only with academic ability, but
‘with several aspects of the student's emotional and mental makeup.
The pertinent question now is the degree to which any or all of
these tests might correlate with the subject student's actual col-
lege performance. Since any one of the above tests or scales could
theoretically correlate well with academic performance, it is neces-
sary to include instruments that do not, on the surface, seem to be

related to academic ability. The IHPI is a good case in point.

III-1

CHHPTER THREE

1 REVIE' OF CURRENT ATTEKPTS TO PREDICT
ACADEMIC PERFORIANCI

The professional literature dealing with attempts to predict
academic success bears a strong resemblance to a literary Irish
stew. It consists of a heterogeneous mixture of predictive vari-
ables, success criteria and methods of Operation with little or no
over-all pattern or organization discernible. There seems to be a
tendency for many of the writers in this area simply to take what—
ever materials are at hand and make a study of them. lore often
than not these studies are attempts to salvage something from edu-
cational testing programs that have been Operating for years, but
for which no data on their predictive validity has ever been
gathered.

Virtually all colleges and universities administer some type
of admissions testing program for entering students. These test
batteries are generally designed to measure academic ability in such
areas as lathematics, Heading, Grammar, Spelling and general verbal
usage. In some instances highly specific tests may be used for stup
dents interested in technical areas such as Engineering and the sci-

ences. The assumption behind these tests is that in some way they
separate potentially good from potentially poor students. This

assumption has in some instances been proven totally incorrect and

IlI~2

at best only partially correct.

{Attempts to predict academic success are being made on all
educational levels, from pro-school to graduate school. The
results obtained seem to be roughly similar on all levels, so a
review of a narrow segment of this literature should provide a rep-
resentative picture of the general field. Since this particular
study deals with prediction on the college level, the literature
reviewed dealt with studies on this level.

The literature research was conducted during the summer of
1965 in the Ohio State University Libraries. The basis of this re-
search consisted of a review of articles on academic prediction
listed in the ngeholcgical Abstracts from late 1960 through early
1965. Of the total number of articles reviewed, forty five were
relevant to this study. The articles and studies cited in this chap-
ter have been selected on the basis of how representative they are
of certain areas of academic prediction.

It should be noted that studies in this field can approach
their material from several different directions, and each direction
places a slightly different emphasis on the materials involved. Per-
haps the most common form of study simply attempts to find statisti-
cally significant correlations between test scores or other variables
and selected criteria of academic success.

A good example of this type of approach is a 1964 study by

6
Michael, Baker and Jones. Using a sample of Freshmen men and women

 

6
William B. Michael, David Baker, and Robert A. Jones, "A.N0te

Concerning the Predictive Validities of Selected Cognitives and Hen-
Cognitive Measures for Freshman Students in a Liberal Arts College,”
Educational and Psychological Measurement, Vbl. XXIV, No. 2, (1964)

373-375. ,

III~3

in a liberal arts college they administered a battery of tests and
worked out correlations with grade point averages (GPA) for their
Freshman year. The tests employed in this study were:

1. California Study methods Survey (CSMS) with four scales -
Attitudes toward school, mechanics of study, planning and
system, and verification;

2. Type Indicator (TI) with four scales - Introversion,
Intuition, Thinking, and Perception;

3. High school GPA's in Algebra and Chemistry;

4. University of Southern California English Classification
Test (nor);

5. Scholastic Aptitude Test (SAT) with three scales - verbal,
Quantitative, and Total.

Of all these scales, CSMS Attitudes toward school yielded the
highest correlations for men (r a .42) while the SAT Total was the
highest correlator for women (r = .48). An examination of Table I
reveals three interesting characteristics. One is that with the
exception of the Introversion scale of the Type Indicator test,
correlations for men and women are either both significant or not
significant. Second, with the exception of the English Classifica-
tion Test, correlations for men and women are within .15 points of
each other and often much closer. Third, the scores for women seem
to correlate better with academic performance than do the scores

for men.

, la
TAnLE I
CGERELKTION OF SELECTED VARIABLES WlTH THE GPA's
0F F‘RESFMEN IN A LIBERAL. ARTS COLLEGE

 

WW“. ”a "AH-Au —--— —-.—‘

 

 

 

.__.a__.. '” Men Timon
Sample Sample
_ Test given size r size r
snr Verbal 210 .35b 177 .44°
Quantitative 210 .24 177 .39
Total 210 .35 177 .48
CSMS Attitudes toward school 155 .42 141 .39
mechanics of study 155 .36 141 .22
Planning and system 155 .33 141 .18
Verification 155 .09 141 .10
Type Indicator Introversion 140 .16 131 .24
Intuition 140 .06 131 .07
Thinking 140 .02 131 .02
Perception 140 .08 131 .00
High School GPA”s in Algebra 161 .06 69 .06
Chemistry 173 .03 113 .16
English Classification Test 210 .23 177 .47

 

 

 

 

 

a
After Michael, Baker and Jones (1964).

b
r's greater than .22 and .17 are significant to the
0173 and 05% levels respectively (when n == 140).
c
r's greater than .302 and .232 are significant to
the 01% and 05% levels (when n = 70).

7 .
Another prediction study was made by Arvo E. deals in 1963

 

7Arvo E. Juola, "The Deve10pment of an Academic Predictor
Scale Based on Students' Attitudes Toward Education," The Personnel
Guidance Journal, Vol. XLII (December, 1963) 381-386.

III-5

based on both ability and attitude measuring instruments. He was
exploring the possibility of predicting academic success on the
basis of attitudes toward education rather than using just ability
measures. In addition to the College Qualification Test (com),
Juola employed the Academic Attitude Preference Inventory (AAPI)
wherein the student rates activities or situations likely to be
encountered in college on the basis of how much he likes or dis-

likes them.

TABLE 23

COMPARISON OF CORRELATIONS BASED ON ABILITY AND ATTITUHES
WITH FIRST TERM AND FIRST FULL YEAR COLLEIB SUCCESS

 

 

 

__L Test Correlations with GPAIs
Tests given Firgt term GPAq Freshman year GPA
men women men women
AAPI .52° .52 .48 .48
CQT verbal .50 .47 .48 .43
Information .53 .58 .49 .56
Numerical .49 .48 .46 .52
Total .63 .65 .59 .64

 

 

 

 

 

aAfter Juola, Arvo E. (1963).
bBased on samples of 212 men and 210 women.
0All r's significant to the .OT% level.

Examination of Table 2 seems to indicate that the AAPI may
have a certain consistency between men and women that the CQT lacks,
at least in Jucla's study. However, since very few instruments
yield identical correlations for both sexes, this may well have been

a statistical "accident". In degree of correlation, the AAPI seems

III~6

to be inferior to both the Information and Total scores of the CQT.
Two other points of interest in the preceding table are that corre-
lations appear to be consistently higher for first term GPAFs than

for the full year, and that correlations with womens' grades again

seem higher than for men. (Could it be that women really are more

predictable?)

An especially relevant study by King, Norrell and Erlandson8
dealt with 375 male Police Administration students here at Michigan
State University and used several of the instruments that are
employed in this study. Thble 3 summarizes the major results of
this study between test scores and first term grades. In general,
correlations seem uniformly low in this study, with only the total
score for the MSU Reading Test and the Language scale of the Calif-
ornia Test of mental Maturity even rising as high as .50. Twenty
eight of the thirty seven correlations are below .25. It is inter—
esting to compare the results of this study to what Michael. Baker
and Jones discovered about the Type Indicator. Both studies seem
to support the contention that tests designed to measure aspects of
the personality (Type Indicator, Minnesota Multiphasic Personality
Inventory and the Thurstone Temperament Schedule) seem to bear
little relation to academic performance. It is also interesting
to note that vocational preference, as measured by the Kuder test,

also seems to have little bearing on the problem.

 

8Paul King, Gwen Ferrell and F. L. Erlandson, “The Prediction
of Academic Success in A Police Administration Curriculum," Educa-
tional and Psychological Measurement, V01. XIX, No. 4 (1958), 649-
651.

IIIIY

TABLE 38

CORRELATION-OF CERTAIN TEST AND SUBwTEST
SCORES WITH THE FIRST TERM GRADES OF
POLICE ADMINISTRATION FRESHMEN

variable Correlation

American Council on Education

Quantitative . . . . . . . . . . .22b

Ling‘liatio e e s e s e e o e e e 043
{Pctaleeeeeeseeeese e38
M.S.U. Reading Test
vocabulary . . . . . . . . . . . .46
comprehension . . . . . . . . . .44
Tbtal . . . . . . . . . . . . . .50
M.S.U. English Test . . . . . . . . . .47
M.S.U. Arithmetic Test . . . . . . . .26
California That of Mental Maturity
Language . . . . . . . . . . . . .50
NOn Language . . . . . . . . . . .17
Tataleeseeeeeeeeee c46
Kuder Preference Record C Form
Outdoor 0 O O 0 O O O O O O O O .02
Mechanical . . . . . . . . . . . .05
computational e e e e s e e e e -s14
Scientific . . . . . . . . . . . .18
Persuasive . . . . . . . . . . . .Ol
”Distic s s e e e e e e e e e e e04
Literary . . . . . . . . . . . . .08
maicaleeeeooesooee’ell
Social Service . . . . . . . . . .04
Cl.rical e e e s e e e o o e e s "oll
Minnesota Multiphasic
Personality Inventory
Hypochcndriasis . . . . . . . . .08
Depression . . . . . . . . . . . .13
Hysteria . . . . . . . . . . . . .05
Psychopathic Deviate . . . . . . .13
Masculine-Feminine . . . . . . . -.Ol
ParanOiaeeeeeeeoeeee e23
SchIZOphrOnla e e e e e e e e e e06
maniaeeeeeeoeessee oo4

Continued on next page

ILIAS
Table 3 wContinued

Variable Correlation

Thurstone Thmperament Schedule
ACtlve e e e e e o e a o e e a ”-005
Vigorous . . . . . . . -.O8
Impulsive . . . . . —.ll

0 O o O
9
e
O
o
J

. . . . .
Dominant . . . . . . . . . . . .00
Stable . . . . . . . . . . . . . .04
Sociable . . . . . . . . . . . . .0.
Reflective . . . . . . . . . . . .20

a ,
After King, Norrell and Erlandson (1959?
r's greater than .148 and .113 are signix
ficant to the .01% and .05% levels respectively
(when N - 300)

Single test or sub-test correlations with college grades are
generally poor, at best ranging around .50 and with large differu
ences between tests. It is also interesting to note that differ-
ent researchers often get entirely different results even when

using the same tests as predictor variables. Table 4 shows, in sums

mary form, the results of three studies correlating the verbal and

TABLE 4

CORRELATIONS 0F FFEfim EN GEA°S WITH TWO .CALES
0F THJ SCHOLASTIC APTITUIE IxfiST

 

 

 

 

 

 

 

 

 

-1... " .1... -:.,h .;.. .1...
Authors of study :AT verbal _m:'Anguant1tative.
Griffen and Flaugherty9 .63** .59**
10 .

Watley and Mervin .48*** .45***
Michael. Baker and11 .35** gmen} .24“ (men) ‘

.ones .44** women; .39*‘ (women)

*

Significant to the .05 level

*‘i .
Significant to the .01 Level

III~9

quantitative scales of the scholastic Aptitude Test with grades for
college Freshmen. This table shows the same predictor variables
yielding significantly different correlations when used with differ-
ent samples of students. This is rather typical of studies in this
area and would seem to indicate that test scores are not the only

variables involved in academic performance.

SPECIALIZED TEST BATTERIES

One line of thought that has involved several researchers is
that perhaps specialized tests should be designed for each distinct
academic area. Underlying this thought is a feeling that trying to
predict the performance of a group of college students majoring in
different areas may tend to cancel out predictive validity for some
areas. That is, if tests are given to students in several academic
areas, very high scores for some groups of students and very low
ones for others could cancel each other out and yield nothing but
an average figure incapable of predicting anything.

Providing at least partial support for the idea of giving dif-

ferent tests in different academic areas is a 1964 study by Lewis,

 

*‘IHI'
No data given on significance.

glary Louise Griffin and Sistem E. Rita Flaherty, ”Correlation
of CPI Traits with Academic Achievement," Educational and Peycholo~
51°31 measurement, Vol. XXIV, No. 2 (1964)-3‘6““‘3’19- 72.

10
Donivan J. latley and Jack C. Kerwin, "The Effectiveness of
variables for Predicting Academic Achievement for Business Students",
The Journal of Experimental Education, (Winter, 1964) 189-192.

1
Michael, Baker and Jones, loc. cit.

111-10

12
Braskamp and Statler. In this study they administered a speci-

ally designed Law School Admissions Test (LSAT) for correlation

TABLE 5“1
GPA CORRELATIONS WITH A SPECIALLY DESIGNED TEST FOR
PREDICTING SUCCESS IN.A COLLEGE OF LA!
Test given ' Correlation
Law School Admissions Test
Total . . . . . . . . . . . . . .39b
writing ability . . . . . . . . .28
General background . . . . . . .18

Undorgraduato GP‘ e e e e e e e e 0 e26

 

aAfter Lewis, Braskamp and Statler
bAll r's over .20 significant to 1%
level (when n - 180)
'with first semester grades in a Law school. (See Table 5). 0f
the four measures employed, the LSAT total did yield the highest
correlation (r - .39), and may indicate a relationship between the
test and Law School performance, but a tenuous one at best.
A 1964 study by Boel3 also gives some support to this idea,

but in a rather indirect manner. He administered a test battery to

Engineering students and then computed correlations with grades

 

12John I. Lewis, Larry Braskamp and Charles Statler, "Pre-
dieting Achievement in a College of Law,“ Educational and Pa cho-
logical Measurement, Vol. mv, No. 4 (196T‘u4 947~'94'°9". ”'""‘z' "'

13Erling E. Boe, "The Prediction of Academic Performance of

Engineering Students,“ Educational and Psychological Measurement,
Vol. xnv, No. 2 (1964) 377-383. ' """ '

III-11

earned in Engineering courses, in non-Engineering courses, and in
all courses combined. Table 6 shows that when the correlations
were broken into two groups they appeared to increase for one
group (non-Engineering courses) and decreased for the other group

(Engineering courses).

TABLE 68

TEST CORRELATIONS‘IITH GRADES IN ENGINEERING
AND NON—ENGINEERING COURSES FOR
ENGINEERING'IAJORS

 

 

 

 

 

Csrrelations
Engineering Other
Tests given courses courses Cbmbined
American Council on Education
Linguistic .209* .626** .358*'
Quantitative .037 .184? .107
I"!
Tbtal .162 .450 .298
as as
Cooperative English Test .282 .561 .430
Wisconsin State College ** ** **
lath Placement Test .344}, .407“ .418
I"! it! {.5
High School GPA .327. .556 .451 ’

 

 

 

 

*Signifisant to .05 level
i”Significant to .01 level

3After Boo (1964)

The greatest change was with the Linguistic scale of the
ACE which appeared to increase by .168 (from .358 to .526) for
non-Engineering courses and decrease by .149 (to .209) for Engin»

eering courses. Two other facts stand out in this table: (1) co-

III~12

rrelations for non-Engineering courses were consistently more sig-
nificant, and (2) the lowest correlation with Engineering GPAuwas,
surprisingly, the Quantitative scale of the ACE. English and Lin-
guistic measures seemed to provide the best correlations for non-
Engineering courses and the specially designed WSC Hath Placement
Test was the highest for Engineering areas. These results would
seem to indicate that different variables are influencing grades
in the different areas, and that special tests may be appropriate.
Emwever, since the operating variables are so difficult to
isolate, not all attempts to predict with specialized batteries
have succeeded. In a 1961 study, Ellis14 arrived at the conclu-
sion that specialized test batteries did not seem to be worth the
extra effort involved in their preparation and administration. He
selected a sample of students majoring in Liberal Arts, Engineering
and Cemmerce; administered a specialized test battery to each group;

and then administered a general test battery to all three groups.

TABLE 7
DIFFERENTIAL PREDICTION IN THREE TYPES OF COLLEGE CURRICULA

 

 

GRA.correlations with

 

Curriculum General battegz §pecific batter: Change
Liberal Arts .58* .61 +.03
Engineering .42 .47 +.05
Commerce .53 .55 +.02

 

 

 

 

*r's significant to .01% levels

14Kenneth Ellis, "How Effective is Differential Predic-
tion in Three Types of College Curricula“, Educational and
Psychological Measurement, ve1 XXI, No. 2 19 1 .

 

III-13

Table 7 shows that the specialised batteries did yield slightly
higher correlations, but at best only increased by .05 for students
in Engineering. Considering these small changes in correlation val-
ues, Ellis would seem to have a valid point in his contention that
specially designed test batteries are not justifiable in the light

of their limited success.

HIGH SCHOOL PERFORMANEB AS AN INDICATOR

In spite of all the work that has been done with psychologi-
cal inventories and tests of academic abilities, the best single
predictor of college success at the present time seems to be high
school performance. The performance criteria may be stated in terms
of high school grades earned or on the student's standing in his
class, but in both instances correlations with college performance
are as high or higher than those obtained through the use of tests
and inventories. Table 8 summarizes the results of four studies
that compared the predictive potential of high school factors and
test results. In each of these studies, high school performance
seemed to provide the highest correlations with college GPAIs.
This is not surprising, as a measure of over-all high school per-
formance would include the effects of many non-academic factors
such as family attitude toward education, other interests, etc.

that would also effect college performance.

MULTIPLE FACTOR CORRELATIONS
Fortunately, it is not necessary to rely exclusively on
single factors when attempting to predict academic success. By

combining scores from several measures and assigning different

.omcq cocoon deem .adst some canvas meanness sense use oomsfinownom Hoonom swan mom m.a oasasoo
.
.eae eon .mess
ma
.eee eon .aefieea
a,
“weary
N .62 .>Hxx .Ho> .psoacnsmcos Hoes cacao mm was Hoscapoosom .emswsamna msfimasz on mmocosm wo oncoofiooam no
meanness Hosanmcnmoam dos odspflpq< .Pscso>ofino<: .omsm .m pooch coo hoaaoz .m such .mmosch .2 snow
we
.eae cos .eem
ma

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

steed so. on eeeesoaemsm ** seems no. oe eeeoaosemsm *
mm. Hcpao>
me. noseesesees page essenen< oneneflosem
mm. moosasnpmmmoge
mm. psflsnpmom canon psossnomsme ossnossfiwlcaomaflso wm. maxscm
om. asapso>sH honeycomxm shag
we. sees :ofieanenaoo sandman mmmo
mo. nonvosospofi
em. Hepne> page essenpne onenesesem we. masses
exme. page see: eweaaoo neeem anemones:
me. ones annsmem eeseeseeoeo
#:0m. fleece
or. o>apcpaascsd .
*xmm. owpmfiswmfia $04 *smv. mw<mc
xsmm. *mm. Amsacsonv psosmoao>cm Hcsoapscscm mo poop HmfipCoSUom *xvm. xxmv. me<mc
soon onEon cams women as concamsc nonsense posse upon oassom case chances
consa.omsom cmoaaoo cons oeaom, mofiaoo Hoosom nmfim
sofipoaoasoo spas.soapsaounno
E

 
  

mflqm<Hm<> mOBOHQmmm mmmeo mEHB Qmm<mfioo «mo 924 ¥z<m doomom mOH:

III-15

weights to these scores one can usually arrive at higher correla-
tions than when single factors are used. The study cited earlier
by King, Horrell and Erlandscn utilised this technique with a
slight improvement in correlations. They combined the language
scale of the California Test of Mental Maturity (r - .50) with the
total score for the I. S. U. Reading Test (r = .50) to obtain an
over-all correlation of .56. The study by Lewis, Braskamp and Stat-
ler (also cited earlier), achieved similar results. In this in-
stance they combined the composite score for the Law School Admis-
sion Test (r = .39) and undergraduate college GPA (r - .26) for a
correlation of .43 with grades earned in the first semester of law
school. Most other studies being reported today also attempt to
combine a number of factors as the basis for their predictions of

academic success.

GEEERAL COMMENTS

Efforts to seriously predict academic performance are rela-
tively recent, and are plagued with many of the problems common to
new disciplines. One big problem area here seems to be a lack of
guidelines as to what constitutes sound study design. Many of the
researchers in this area seem to have a tendency to study whatever
material comes to hand, regardless of how well this material may be
suited to quantitative study. Another problem lies in the report-
ing of these studies in that guidelines are again lacking and that

some writers seem to omit almost as much data as they include.

Hany of their reports are sketchy at best, often leaving out import-

ant information on sample size, statistical significance of find-

III-16

ings and research methodology. A writer may claim a correlation
of .85 or more between some test and academic achievement, but if
he fails to explain how he arrived at a particular figure his
results can not be taken too seriously.

In an area as new and as lacking in background knowledge as
the field of psychological testing, even negative results are valup
able. They at least say something about what is not so, provided,
of course, that they are reported in such a manner as to inspire
confidence. In contrast, the poorly reported or documented study
really says very little, no matter what it may claim to say. For—
tunately there have been enough well documented studies so that
some general comments can be made concerning the ”state of the
art".

If we assume that the published literature reflects develop-
ments in the field of academic prediction, several general conclu-
sions can be drawn.

1. Tests measuring academic ability seem to give higher

correlations than do psychological inventories.

2. Those tests that measure a student's linguistic pro-
ficiency are better predictors than those based on
quantitative skills. This is often even true for
areas such as engineering and the sciences that tend
to rely heavily on quantitative skills.

3. Specialised test batteries seem to do a slightly better
job of predicting college success than do general bat-
teries when given to homogeneous groups.

4. High school performance is as valid an indicator of

 

III-17

academic potential as most test batteries being admin!

istered today. This remains true even‘when high school
performance is compared to scores on tests specifically
designed to predict academic success.

5. lbmen may be more easily predictable than men (in rela-
tion to college performance).

6. If two different people or groups study the same test or
batteryof tests they are likely to get significantly
different results.

7. lb really do not know enough about the prediction of
academic success to be thoroughly satisfied with the
outcome of much of the work that has been done in this
field. The best that can be said is that high school
performance and some test scores do, in some way, seem
to be related to academic performance. However, the
relationship is far from perfect.

In consideration of these shortcomings, one must be extremely
cautious in attempting to predict academic success. Actually,
simply trying to predict success presents about as many problems
as it does results. It is in the use of these correlations that
caution is necessary. Since correlations between test variables
and academic performance are generally weak, there will be numer-
ous exceptions. Iith our present knowledge, any attempt to base
college admission procedures solely on admissions tests or high

school performance would result in many good students being re-

Jected and many poor ones admitted. There are simply too many

variables operating that cannot even be identified.

F*_-,

III-18

l’atley19 has a particularly graphic way of illustrating the
problem involved in this imperfect relationship. He took 109 suc-
cessful (above a 2.00) and 39 unsuccessful (below a 2.00) first
term college freshmen and plotted cumulative percentage curves for
their scores on the Guilford-Zfimmerman Restraint Scale. Table 9
shows these curves. Each point on a curve shows the percent of
students in that category with scores below the corresponding per-
centile on the Restraint Scale. The reason for constructing these
curves was to illustrate visually the effects of establishing an
arbitrary admissions cut-off on the basis of such a test. For
example, if the thirtieth percentile was chosen as the lowest
acceptable score, the dotted lines at the lower left of Thble 9
show that almost thirty percent of the unsuccessful students would
have been rejected. However, this same cut-off point would also
have rejected almost ten percent of the successful students. Since
the curves are so close to parallel, any arbitrarily established
cut-off point will eliminate some of both types of students and
allow some of both types to remain.

lhile correlation of the trait with GPAfiwas only .36, it
still illustrates why caution.must be exercised in the area of
academic prediction. Any correlation below a perfect 1.00 would
also have this tendency to eliminate some good students and admit
poor ones.

In spite of the apparent confusion in this field of predic-

tion, we do hays many good indications of what is or is not so,

 

19
'htley, loo. cit.

 

 

III-19

TABLE 9
munvr mucus canvas ma scams or succsssm
um unsuccsssm 3mm on m 3331mm! scam
100 cu
9° '7'
80 --

unsuccessful

70 -- successful
60 --
so __
40 -P
30 r.
20 -~ I

 

1°, --.__|

/'. ‘
, L 1 s m a J j l
l I 1 j I j 1 T 1

_10 20 30 40 50 60 70 80 90 100

Percentile on Restraint trait

and have established some guidelines for continuing research.
This review of the literature does, however, point out several
areas that are very much in need of improvement. One of these
areas is what one might term the spontaneous study where the ”re-
searcher" seems to study whatever lies at hand simply for the sake
of studying it; and perhaps to get his name in print. There even
seems to be a standard recipe for this type of study. -
”Start with an existing batch of test results given and-I
scored by someone else; fold in an equal amount of suc-
cess criteria; strain through an IRI sorting'maohine;

and then bake in the largest and.most impressive com-
puter available. Finally, analyse results. Caution:

’.~_"f.‘

III-20

do not analyse too critically.”2°

The need is for more studies that are carefully planned from
the beginning of the testing program. In this way research con-
trols can be placed on any factors that may tend to bias or inval-
idate the results. In addition, there should be more unified and_
coordinated research programs in this area. ”Hit or miss” research
programs may well add to our knowledge, but they do so in such a
disorganized manner that the value of their contributions is mini-
mised.

Finally, when research is reported, it should be reported

completely. A.statement of "results” or ”findings" is virtually

meaningless unless it is accompanies by other information support-

ing the findings in such a manner that they can be accepted by

other persons working in the same area.

 

201‘ Olde Reoi e Book for Graduate Student Research Pro-
gectsé (formerly titled: Fort: Fast lays into Print and Higher

1 Positions).

U

IV-1

CHAPTER FOUR
STUDY DESIGN

PRILIIIHIRY PROBLHIS

This particular study, like so many in the area of academic
testing, is something of a compromise measure. The testing program
concerned was not, apparently, established with this type of study
in mind. The full length of the testing period was from the fall
term of 1953 through the fall of 1964. During this period about
1,200 students were actually tested. The final sample for this study,
however, consists of only 277 students representing a four year span
of time.

There were several reasons for this selection of usable test
records. The primary one was a major change in the composition of
the test battery., From its beginning in 1953 through 1958 the bat-

tery consisted of the following tests:

lichigan State University Rngrance Examination (American
council on Education Psychological Examination)

Inder Preference Record - Occupational 0 Form

linnesota Hultiphasic Personality Inventory

California That of Mental laturity
580 students took at least part of this test battery, and 466 of
them had complete sets of scores. A.signifioant complication was
that many of the available scores were not reported in a manner

suited to correlation analysis. To obtain a meaningful coefficient

IV-2

of correlation between two variables, there must be a fairly large
range within which each variable can fluctuate. For example, grade
point average could fluctuate from 0.00 (failed all courses) to 4.00
(perfect), while a test score reported in percentiles could vary
from 00 (bottom) to 99 (top).
unfortunately, many of the scores during this period were
reported in terms of rather narrow categories. For example, scores
for the California Test of Isntal laturity were given only as "high“,
"medium high”, ”medium", ”medium low”, and "low”. The Kuder Prefer-
ence Record was also, unfortunately, scored in terms of "high”, "med-
ium”, and "low” for each scale. Since the actual scores that the
students obtained on these tests were in many instances no longer
available, this first test period had to be disregarded.
During the second testing period the following tests and
scales were used:
lichigan State university Entrance Examinations
College Qualification Tests with verbal, numerical, gen-
eral information and total score scales
l. S. U. Reading 'Best
I. S. U. English Placement Test
Otis Test of Mental Ability
linnesota Iultiphasic Personality Inventory (3 validity
scales and 10 clinical scales)
Kuder Preference Record D Form
Strong vocational Interest Inventory
‘1 total of about 600 students took at least part of this battery of

tests but several considerations dictated a reduction in the number

 

IV-3

of records used. One reason for this was that the records for each
student had to be complete for every test, so any records with mis—
sing scores were discarded. Another reason was that the study re-
quired full data on the subjects' performance in college. Because
of this requirement no tests given after the fall of 1961 could be
used since the later groups would not have completed their univer-
sity program when this study began. In addition to a reduction in
sample size, the finder Preference Record and the Strong vocational
Interest Inventory had to be eliminated because of the same quali-
tative scoring that necessitated the rejection of the earlier half
of the testing program. Finally, since this study deals with aca-
demic prediction in a Police Administration curriculum, all test

records for students not in Police Administration were rejected.

TH! PIHAL SAIPLH

The final sample consisted of test records that met the fol-

lowing criteria

1. The scores for the various tests had to be reported in
a quantitative manner, as raw scores or percentiles.

2. Each record used had to have scores for the I. S. U.
Entrance examinations, the IHPI and the Otis Tost of
lental Ability.

3. The student must have taken the test battery no later
than the fall term of 1961 to allow him to graduate
prior to the beginning of this study.

4. The student must have graduated with a degree in

Police Administration and Public Safety.

IVr4

The group of test scores finally selected for the study consisted of

277 sets of records spread out over the following terms:

TABLE 10

NUHRER OF SUBJECTS 'ITH COMPLETE
TEST SCORES IN EACH TERI

 

 

 

 

"1fififiifi?ifi’
Tbrm Year Sub1=2£3_
Fa11 1958 61
Winter 1959 15
Spring 1959 10
Fall 1959 90
iinter 1960 16
Spring 1960 15
linter 1961 16
Spring 1961 11
Fall 1961 L
' Total 277

 

 

 

This represents a sharp reduction from the ideal of including the

records of every student taking the tests, but is the largest

sample that can be employed in this study.

It should be noted that

this sample includesall students in the School of Police Adminis-

tration and Public Safety for which sufficiently detailed records

‘were available. The sample was not limited to students who suc-

cessfully completed college. In fact, some of the students failed

to complete more than one term.

IV-S

CRITERIA OF ACADEMIC SUCCESS

Selection of measures of academic success was limited by two

major factors. The first of these was inherent in the design of

the study.

Those measures utilized had to be suitable for corre-

lation analysis in that they would have to be capable of varying

over a fairly wide range of scores. They also had to be of such a

nature as to not require too much office staff time to obtain from

the student records.

The success criteria finally settled upon were three:

1.

2.

3.

All university grade point average (GPA). This consists
of the student's total number of honor points divided
by credits earned. An "A" is worth four honor points,

a “H" is worth 3, "C" is worth two, “D" is worth one,
and "F” is worth zero. In this manner a four credit ”A”
would be worth 16 honor points (four credits x four
points - 16). Similarly, two three credit ”B's" and two
three credit "C's" would yield a GPA of 2.50.

Grade point average in Police Administration courses.
This figure is computed in the same manner as the all-
university GPA except that only courses taught in the
school of Police Administration and Public Safety were
counted. It was hoped that this figure would provide a
measure of just how well the student performed in course
work directly related to his chosen profession.

Average number of credits carried per term. This figure
was obtained by dividing the total number of credits

earned while in the university by the number of terms

 

IV—6

spent earning these credits. The reason for including
this figure was to provide added meaning to the all-
university GPA.by making allowances for belowbaverage
students who could only maintain passing grades by
taking light credit loads or above average students

who carried exceptionally heavy study loads.

RECORDING PROCEDURES

The recording of information relative to academic success was

performed by the staff in the office of The School of Police Adminis-

tration and Public Safety. The data were computed and entered on the

reverse sides of the test score report forms in the following manner:

1.

2.

3.

The student's all-university GPA is an official figure com-
puted by the university. The source for this figure was
the "Report of Standings" issued by the university for
each student‘s final term in college.22

GPA in Police Administration courses was computed by list-
ing all Police Administration courses taken along with the
credit values and grades for each course. Dividing total
honor points in these courses into the total number of
Police Administration credits carried gave the desired
measure.

Computing a student's average credit load per term was
simply a matter of dividing the total number of credits

taken by the number of terms spent as a full or part-time

 

22The "Report of Standings" is issued by the University for each
student at the close of each term and shows courses taken, grades
earned, and cumulative totals, including all university GPA.

IVs?

student here. This figure does not include any credits

for courses taken at other schools.
As the material was recorded it was periodically checked against the
files by the writer and one of the office staff.

The final recording step was to enter all data, test scores
and success criteria on IHI forms to serve as a guide for the card}
punch operators at the lichigan State university computer center.
The information entered on the IR! forms was checked for accuracy by

the writer before it was sent to the computer center.

STUDY DESIGN

The study was divided into three major steps, each involving
about one-third of the total sample (see Thble 11) and each step
requiring processing:

1. Run a correlation analysis between the test scores for
the first group (H - 1/3 of total population) and the suc-
cess criteria. Then identify these test or sub-test vari-
ables in the test battery that seem to correlate well with
the criteria. Rfliminate items with relatively low corre-
lations (r's).

2. using the surviving items from step one, and the second
one-third of the population, develop a multiple regression
equation for predicting academic performance from the
scores.

3. 0n the basis of the regression equation obtained in step
two, and using the test scores of the third group, predict
how well the third group will do academically. Run a

correlation analysis between predicted and actual per-

IV-8

formance for the third group.

If the correlation between predicted and achieved performance
for the third group is of the same magnitude as the multiple corre-
lation from the regression equation in step two, the test scores can
be thought of as having a consistent relationship with academic per-
formance. The size or importance of this relationship will be indi-

cated by the size of the relative multiple correlations for the two

groups.

IHPLHHHNTING THE DESIGN

6 Once the general outline for the study had been decided upon
and the sample selected, the actual implementation was a fairly
straightforward matter. The first step was to separate the sample
into three equal parts, using a table of random digits. This was
done by numbering the test scores in each term from one to n. In
the first term there were sixty sets of test scores. The first
thirty sets of test scores with serial numbers corresponding to the
first thirty random digits were withdrawn and labeled Group A. The
second thirty sets corresponding with.the second thirty random
digits was labeled Group B. Group C consisted of the remaining
thirty sets of test scores. This same procedure was applied to the
remaining eight terms. To further insure random selection of the
scores used in the three computational steps, the table of random
digits was also used to select the relative order in.which Groups A,
R and C would be taken from each term. The final arrangement of the

samples is shown in Table 11.

IV—9

TABLE 11

DIVISION OF TERM SUB-GROUPS INTO THE
THREE COMPUTATIONAL STEPS

 

 

 

Number of Test Records And

Relative Orderin in Each Ste
Term N Step I Step 2 Step E

Fe11, 1958 61 19 (B) 23 (c) 19 (1)

 

 

vinter, 1959 15 4 (c) 6 (B) 5 (1)
Spring. 1959 10 3 (0) 4 (A) 3 (B)
Fan, 1959 90 29 (c) 31 (4) 30 (B)
Tinter, 1960 16 6 (1) 6 (c) 4 (B)
Spring. 1960 15 5 (o) 6 (a) 4 (B)
Winter, 1961 16 8 (H) 4 (c) 4 (1)
Spring, 1961 11 1 (1) 6 (B) 4 (c)
Fall, 1961 43 11 (A) 16 (B) 16 (c)
Total 277 86 102 89

 

 

 

 

 

The next step was to obtain the success criteria from each
student's records and enter them on the forms reporting the test
results. Once this was finished the data could then be arranged
and itemized for computer processing. The variation of totals in
the three steps in Table 11 is a result of missing information in

the office files.

COMPUTER PROGRAM
The computation of correlation coefficients was done on the

University's Control Data 3600 computer as part of a standard ”pack-

IV—10

age" of services available for student research projects. In addi-
tion to showing correlations between all variables on the first two
runs, the computer output also provided means, standard deviations

and the sum of each variable over all observations. The third com-
puter run supplied single and multiple correlation coefficients and

certain information relative to statistical significance.

V~2

Table 12 — continued

 

 

With A11-
__ Univ. GP; with P’LJULA_
variable Mean r sig. r. sig.
6. Information 45 . 9* . 21 5% .14
7. Numerical 40.4* .09 .09
8. Total 44.0.. .22 5% .16
9. MSU English 40.6* .24 5% .18
10. MSU Reading 45.5 .41 1% .35 1%
11. mm L 4.0 —.11 -.05
12. F 3.9 -.12 -.12
13. K 17.0 .12 .04
14. Rs 12.5 -.02 —.13
15. 1) 17.2 -.04 -.10
16. Hy 20.1 -.01 -.O6
17. Pd 22.5 .10 .05
18. 141? 23.2 -.11 -.08
19. Pa 9.0 —.20 -.17
20. Pt 26.3 -.14 —.23 5%
21. So 25.6 -.09 -.16
22. Ma 20.1 -.13 -.04

 

 

 

 

 

 

aSpace left blank indicates significance level below 5%.

*Percentile scores based on total number of entering students
who took the M. S. U. entrance battery in a given term.

The most obvious thing about Table 12 is that none of the
independent variables seem to correlate particularly well with aca-

demic performance. For prediction purposes, the MMPI was a dismal

CHAPTER FIVE

RESULTS OF STUDY

THE FIRST COMPUTER RUN

The first computer run was based on data for eighty six sub-
jects and compared twenty two variables for each subject. The
dependent variables were the two grade point averages while the
independent variables consisted of the various test scores and the
average number of credits carried per term. As was mentioned in
the preceding chapter, the primary purpose of the first run was to
reduce the number of variables by eliminating those bearing no
strong relationship with academic performance. Table 12 shows the

computer output relevant to this phase of the study.

TABLE 12

AVERAGES 0F VARIABLES AND CORRELATIONS WITH ALL—UNIVERSITY
AND POLICE ADMINISTRATION GRADE POINT AVERAGES

 

 

=========================F=========F====IFiEEEF====================
Uiiv. GPA With PLA GPA
---Vhriab1e Mean, r EEZQ_ r IFSTE:::
1. 111 University an 2.37 .89 1%
2. PLA GPA 2.48 .89 1%
3. Credits per term 14.6 .30 T% .25 5%
4. Otis IQ score 113. .20 a .17
5. CQT verbal 47.1* .18 .12

 

 

 

 

 

CHAPTER FIVE
RESULTS OF STUDY

THE FIRST COMPUTER HUN

The first computer run was based on data for eighty six sub-
jects and compared twenty two variables for each subject. The
dependent variables were the two grade point averages while the
independent variables consisted of the various test scores and the
average number of credits carried per term. As was mentioned in
the preceding chapter, the primary purpose of the first run was to
reduce the number of variables by eliminating those bearing no
strong relationship with academic performance. Table 12 shows the

computer output relevant to this phase of the study.

TABLE 12

AVERAGES OF VARIABLES AND CORRELATIONS WITH ALL-UNIVERSITY
AND POLICE ADMINISTRATION GRADE POINT AVERAGES

E

Univ. GPA With PLA GPA
r

 

 

'"“'V‘erieble Mega r Big 315:.
1. All University on 2.37 .89 1%
2. PLA GPA 2.48 .89 1%

3. Credits per term 14.6 .30 1% .25 5%
4. Otis IQ score 113. .20 a .17
5. CQT verbal 47.1* .18 .12

 

 

 

 

 

IJ... 3

failure with most of the correlations clustering around zero. The
large number of negative correlations between the various scales of
this instrument and academic success is to be expected since high
scores on most MMPI scales could be indicative of mental or emotional
problems. most of these problems would probably reduce a student's
effectiveness. It should be noted that, aside from a larger number
of negative correlations, these findings on the MMPI tend to cor-
respond with the findings of King, Norrel and Erlandson. (See

Table 3). On the basis of the results of this first computer run,
the MMPI was dropped from further consideration.

The highest correlation in the first run was .89, but this
simply deals with the relationship between the two grade point cri-
teria and has no predictive value. A student who does well in one
area of academic study can generally be expected to do well in
other areas also. Average credits per term show a significant
positive relationship with academic performance, but being an
"after the fact" calculation, cannot be used for prediction. In
addition, neither of these correlations represents a cause and ef-
fect relationship. Rather, the pairs of items show a somewhat con-
sistent relationship because they are influenced by the same out-
side factors. For example, while the better students will tend to
carry heavier course loads, these course loads do not determine
their grades. Their grades are determined by generally superior
scholarship.

According to the results of Table 12, the most promising
instruments seem to be the specially designed MSU English and Read-

ing tests. Their correlations with the criteria were .24 and .41

 

V—4

respectively, for all-University Grade Point Averages, and .16 and
.35 for Po1ice Administration Grade Point Averages. Only the MSU
Reading That was statistically significant for both GPA's. Both

the CQT and the Otis That of mental Ability yielded disappointingly
low correlations with the criteria. They do seem to bear some re-
lationship to success, but only the CQT Information and Total scores
yield statistically significant correlations with the success cri-
teria; and then only in relation to all-University grade point aver-
age.

The correlation matrix shown in Table 13 gives the intercor-
relations between all variables except the various MHPI scales.
These were eliminated from this table because they fail to corre—
late significantly with anything but themselves. It should be
noted that, on the whole, the tests correlate better with them-
selves than with either of the two success criteria. For example,
variable number 8, the CQT Total, shows correlations of only .22
and .16 with the GPA's. However, when the total is compared to the
sub scores of the CQT; namely verbal, Information and Numerical,
the correlations are .80, .78 and .63, respectively. A comparison
of the CQT Total and the Otis IQ Test shows the same type of fairly
high relationship. The correlation between two measures is .66,
much higher than either measures correlation with the success cri-
teria. From this it would seem that the tests are doing a better
job of measuring some other factor than they are of measuring aca-
demic performance. It appears that the tests are measuring aca-
demic mental ability and achievement, but that there is more to
succeeding in college than simply possessing mental ability and

prior academic achievement.

 

.hae>apoonmoa

.oaobea P0. was no. 0:9 09 vceoamaswfim one mmm. one firm. can» neveehm e.n

 

 

oo.H am. pm. me. on. mm. me. me. mm. he. muaeamm 2mg .0?
oo.H mm. av. mm. mm. mm. we. we. em. nmaawnm pms. .m
coca mm. m». om. we. om. we. mm. Happy .m
oo.H mm. we. be. «P. mo. mo. Haofinoaaz .>
oo.H .mv. me. No. er. Pm. cofipuanoocH .w
oo.H an. we. NP. we. Hanan» ado .m
00.? P.. we. om. pmop aH nape .e
oo.H mm. cm. upaeono ommnoaq .m
00.? mm. «me «an .m
00.? «mo apamno>asp HH< .P
.IIIIkaIILum m, m] m n v m [rm F

 

 

 

 

oanmanm canmanm>

 

mom mmabmsbo 9mmHh
mmmao muda_maHB and
maidmahd BZHOQ andmc ODE HEB
mBHB mmndeM4> nmaonqflm ho mZOHadqmmmoo

m9 Handy

THE SECOND COMPUTER RUN

It was decided to eliminate the MMPI and average credits per
term from further consideration for reasons indicated above, and to
base the second run on the Otis Test, the CQT scores and the M.S.U.
English and Reading tests. This second phase of computer work was
based on a sample of 102 students. In addition to developing re;
gression equations, the second-phase computer output also included
single item correlations between all variables as in the first run.

Unfortunately, the second phase results did not match the wri-
ter's expectation of what they should have been. The idea behind
random selection of the three groups was to insure, insofar as pos-
sible, equal distribution of talent among the three groups. If the
equal distribution of talent and other relevant factors could be
achieved, one would expect the correlations within the three groups
to be stable from one computer run to the next. Table 14 clearly
shows that the correlations between test scores and success cri-
teria within the first and second groups were anything'but stable.
All correlations between test scores and criteria in the second
group were higher than in the first group. In some instances the
increases were minor, but many correlations more than doubled in
numerical value from the first to the second run. While the right
hand column in Table 14 shows that only two of these increases are
significant to the .05 level, the fact still remains that all of
the correlations increased by at least a few points. (Significance
data were derived from a table of Fisher's Z.)

While it is remotely possible that these increases may not be

significant, a more likely assumption is that bias was somehow

TABLE 14

V-7

A COMPARISON OF SELECTED PORTIONS OF

11

THE RESULTS OF THE FIRST AND

SECOND COMPUTER RUNS

 

 

Correlations with all—University GPA

 

‘Eggiable First run Second run Increase 2*
Otis IQ Test .20. .30 .10 .718
CQT verbal .18 .41 .23 1.705
Information .21 .45 .24 1.825
Numerical .09 .39 .30 2.16 *
Total .22 .53 .31 2.464*
M.S.U. English .23 .30 .07 .506
M.S.U. Reading .41 .47 .06 .501
Correlations with PLA GPA
Otis IQ Test .17 .22 .05 .350
CQT verbal .12 .33 021 1.490
Information .14 .35 .21 1.500
Numerical .09 .30 .21 1.470
Tbtal .16 .41 .25 1.84
M.S.U. English .18 .23 .05 .350
M. S.U. Roading . 35 .38 .03 . 23

 

*When Z is equal to or greater than 1.96,

significant to the .05 level.

difference is

introduced into the sampling and sorting procedures early in the

study; A.comparison of the correlation matrices for the two runs

(Tables 13 and 15) supports this assumption.

The reader will note

.haebfipcommon

.mHo>oH F0. was mo. on» o» psdofimwsmdm ohm vmm. use mar. some Hepoonm m.H

 

oo.a mm. me. an. mm. as. an. mm. as. anaesom .p.m.s .m

oo.H mm. «a. mm. we. we. mm. on. sandman .p.m.2 .w

oo.H am. am. ma. we. av. em. Heaps .>

oo.H am. om. me. on. mm. asoanoasz .m

as oo.H me. mm. mm. ms. soapsanomcH .m
w 84 E. 2. :1 H8...» So é
oo.H mm. on. same 0H mapo .m

oo.H om. «as «am .m

oo.H «as apamao>asp Han .H

mbaenspwr.

moapmfinm> A capmflnm>

 

25m mmabmzoo 9200mm
mmmeo mu<m NEHB 92¢ mmc<mm>< ezHom
mfi¢mc 0&9 fine mBHB mfiqdem¢> nmaomqfim mo mZOHadgmmmoo

mw mnmda

V-9

that while test score-GPA correlations increased, the inter-corre-
lations between test variables do not vary to any appreciable degree.
While some of them do vary slightly, the variations seem to be small
and random. This would seem to indicate that whatever bias may
exist in the samples is primarily related to success factors, and
not to the test scores. Exactly what this bias is, or how it opera-
tes, is unclear. In fact, there is no assurance than any bias even
exists. However, the likelihood of accidental bias was such that
certain portions of this study were repeated. Before examining the

‘ results of this replication, we would do well to return to the sec-
ond and third stage results of the original study plan.

The reader will recall from Chapter Three that the primary
purpose of the second run was to deveIOp regression equations for
predicting academic success on the basis of test scores. These
equations were developed by an automatic computer process that
screened out test variables that failed to contribute significant-
ly to the equation, and retained those that did. In this particu-
lar instance, the criteria for including any given test variable
in the equation was the statistical significance of the variable
when compared to one of the success criteria. Theoretically, the
regression equations could have included any number of the seven
test variables found to bear a significant relationship to the suc-
cess criteria on any reasonable level of statistical significance.
(The .001, .005, .01 and .05 levels of significance were tried for
all seven variables against both success criteria.)

Since only the CQT Total seemed to be significant, the re-

gression equations were calculated for just that one variable.

V410

The formula used was:

= £1 _ §1-—
Y I'xy (8x X r'xy <Sx> X + Y

2 2
2x Sy = 2X
Nx-l Ny-l

and 2x2 = 2x2 _ Léfili 2Y2 = 2Y2 _ (£322

 

where Sx

Y refers to the predicted GPA

rxy is the correlation between CQT Total

and one of the success criteria
When the actual values were inserted into the formulas, the equa-
tions evolved as follows:

for predicting all-University GPA

Y a .016 (CQT Total) + 1.49

for predicting PLA GPA

Y = .01 (CQT Total) + 1.89

CROSS VALIDATION

The second computer run and related statistical manipulations
yielded a regression equation for predicting academic success. The
purpose of cross validation was to see if this equation, derived
from one sample of students, will successfully predict the perform-
ance of another sample on the basis of their own test scores. The

first step in this process was to use the regression equations

ve11

derived above to predict all-University and PLA GPA's for the third
group.

After computing the predicted GPA.for each student in the
third group, the following formula was used to compute the correla-

tion between GPA's predicted and actually achieved.

N£(XY) — (2x) (XY)
(Nzx2 — [XX]2) (NzY2 — [zY]2)

 

Table 16 shows the results of this cross validation in sum-

mary form.

TABLE 16

CROSS VALIDATION OF REGRESSION EQUATIONS:
R's BETWEEN PREDICTED AND ACHIEVED
GPA's AND Z VALUES OF CHANGES

 

—.

 

 

 

 

All UhiversitzZEEiEFifiir7fiir-
Group 2 .534 .412
Group 3 .427 .273
Z 0.89 1.08

 

 

 

Z of less than 1.96 indicates
changes are not significant.

When subjected to cross-validation, the predictive efficiency of
the regression equations seemed to be reduced for both GPA's. The

correlations between predicted and achieved all-University GPA
dropped from .534 to .427, while the PLA correlation drapped from

.412 to .273.

V-12

From this it would seem that a student's total performance

on the CQT'bears a consistently significant relationship with both
his over-all academic performanceand his performance in Police Admin-
istration courses. The relationship seems to be stronger for over-
all university performance than for the portion of this performance
represented by the student's grade point average in Police Adminis-
tration courses. The actual importance of this relationship will be
explored in detail in Chapter VI, after we have covered the addition-

al findings resulting from re—running portions of this study.

CROSS CHECKING FOR BIAS

Since it seemed highly possible that the original sampling
procedures in this study may have been biased, it was decided to run
some cross checks on the extent of any possible bias. The first
step was to recombine all the usable test records for a new divi-
sion into two groups. The procedure here was to assign random num-
bers to the test records in each term, and then to pick half of each
term's records through the use of a table of random numbers. This
process resulted in two groups of test records (N = 140 and 141) with
an approximately equal number of test records from each term.

For the first group (N a 140), correlations were computed.be-
tween CQT Total scores and the two grade point criteria. These cor-
relations turned out to be .651 for all-University GPA and .307 for
PLA GPA. The formula shown on page va1o was then used to compute
regression equations for predicting the two GPA's for students in
the second group (N a 141). The final equations were:

for all-University GPA

Y = .053 (CQT Total) + .08

va13

for PLA GPA
Y = .037 (CQT Total) = .86
The results of this cross checking are shown, in summary form, in

Table 17

TABLE 17

SIMPLE CORRELATIONS BETWEEN CQT TOTAL AND THE TWO GPA's
AND RESULTS OF CROSS VALIDATION OF
THE REGRESSION EQUATION

1 Correlations wit:

' u" '1' Un"'i"v' ers'i' ty' CPL PLA GPA

CQT total and
Academic Performance

First Group .651 .301

Predicted and
Achieved Performance
Second Group .5454 .281

 

 

 

 

2 ML 0.27

 

As in Table 16, the correlations drapped somewhat when subjected to
cross validation. The Z values of these changes are low enough to
warrant the assumption that these drops are not statistically sig-
nificant. Again, it appears as if a student's total score on the

College Qualification Test is at least partially indicative of how

he will perform academically in college.

VI-1
CHAPTER SIX
ANALYSIS OF RESULTS AND CONCLUSIONS

COMPARISON WITH OTHER STUDIES

By now the reader may have decided that the results of this
study are little different from the results of most of the studies
cited in Chapter III. In one respect at least, this is quite true.
The relationships revealed by this study are far from strong and
only explain or account for a small portion of the factors influ;
encing academic success, as discussed below. However, the cross
validation built into this study insures that a greater degree of
confidence can be placed in these findings than in the results of
most correlation studies concerning academic success reported in
the literature. Host of the writers in this area seem dontent sim-'
ply to report correlations between variables and, evidently assume
that their discovered relationships will held constant. very few
writers seem to make any effort at cross validation.

In this study we have seen that many of the correlations be-
tween test scores and success criteria do not held constant. Howb
ever, we have also seen that this fluctuation does not destroy the
predictive capabilities of the test measure studied.most closely —
the CQT Total score. And. finally, the cross validation procedures
in this study (done twice, actually) leave one with a feeling of

relative confidence in these results.

THE MEANING OF THE STATISTICAL RESULTS
Several of the test scores studied seem to bear some relation-

ship to the academic success of Police Administration majors at

VI—2

Michigan State University. The most notable among these are the
specially designed M. S. U. English and Reading tests, and the
total score for the College Qualification Test. Other test meas-
ures, such as the MMPI, seem to have no significant bearing on the
subject.

After studying the CQT so thoroughly, the question now arises
as to the practical significance of these findings. During the
course of this study, correlations between the CQT Total and all—
University GPA.ranged from a low of .22 to a high of .65. One way
of looking at the meaning of a correlation is to square the figure.
This new figure is an estimate of how much of the variation in the
dependent variable is accounted for, or associated with variation in
independent variable. For example, the above correlations of .22 and
.65 indicate that variation in the CQT Total score (or more pre-
cisely, the academic ability measured by this score) accounts for,
or is associated with, from 5% to 42% of the variation in all-Uni-
versity grade point averages.

In summary, it would seem that the level of intellectual
ability indicated by a student's total score on the College Qualifi-
cation Test consistently bears a small but fluctuating relationship
to his academic performance. This relationship is stronger for all-
University grade point average than for Police Administration grade
point average. That this relationship is small is not surprising
when one considers the multitude of variables that can effect aca-
demic success. Just a few of these are intelligence, students'
attitude toward education, family attitudes toward education, health

and personal problems. Unfortunately, we are in no position to say

VI-3

how or to what extent these other variables may be operating in any
given student. And, finally, as was discussed at the close of Chap-
ter III, any attempts to use test scores such as those studied here
as screening devices for admission should proceed with extreme cau-
tion. There are too many other variables operating about which we

know all too little.

EVALUATION OF THE TESTING PROGRAM

In general, the idea of gathering and studying data relative
to academic success in college is a good one. However, if these
studies are to be effective, to yield.meaningful results, the stud?
ies must be well planned and carefully implemented. The former
testing program in the School of Police Administration and Public
Safety fails to meet these requirements. There are several reasons
for this, and the most important of them lies in the area of long
range planning. When the study was first prepared, its originators
did have certain ideas on what the study should accomplish. They
were particularly interested in exploring the personality and inter-
est makeup of entering Police Administration students. There was
also, apparently, some thought given to isolating tests with pre-
dictive value. Unfortunately, and partly because of personnel turn-
over, no single person was ever given continuing responsibility for
supervising the program. As a result of this, the program responsi-
bility passed through the hands of a series of disinterested, or at
best moderately interested but untrained, custodians. Some students
never took the tests, some scores were misplaced, and even the scor-
ing procedures were changed several times. There was some short-run

use made of the individual results for counseling and advising pur-

VI—4

poses, but there seems to have been no plan for over-all analysis.
This made any study of the whole program virtually impossible.

This is not to say that exploratory testing programs should
not be conducted in the future by the School of Police Administra-
tion and Public Safety. In fact, they should be; but according to
some sort of over-all plan. There is no reason that such studies
could not serve short term advising and counseling needs as well as

accumulating data for intensive analysis.

SUGGESTIONS FOR FUTURE STUDIES

One would hope that the School of Police Administration and
Public Safety would extend its leadership role in law enforcement
education into the study of the effects of this education and ways
to improve its quality. With careful planning and implementation,
there is every reason to believe that such studies would.make sig-
nificant contributions to our knowledge in this area.

However, continued research into the relationship between
academic ability and academic success may contribute less than
studies will in certain other areas. It may well be that studies
of this area have reached a temporary "dead end" until such time as
we learn more about some of the other factors effecting academic
success. Some of these "other factors" of particular interest are
the student's emotional stability and.maturity, his attitudes to-
‘ward education, his occupational and vocational interests and his
family—educational—cultural background. An area that might be of
particular interest to the School of Police Administration and Pub-
lic Safety is the personality make—up of students in this field.

we ought to be concerned with exploring the nature of the people

VI-5

'who are attracted to careers in law enforcement and related areas.
There are presently in existence test devices or procedures that
'will at least estimate the extent of these various factors in a
student. Studies in these areas would require time and money, but
would not be of prohibitive difficulty. This would hold especially
true if the School of Police Administration and Public Safety
availed itself of the consulting services available from both the
Counseling Center and the Psychology Department prior to initiating
further studies.

Finally, there is one area that may well be highly important,
but is seldom studied That is the relationship between academic
ability, academic success, the "other factors" mentioned and the
success and job satisfaction of the student after graduation. This
type of study would.be very long range, perhaps covering ten to
twenty years in all. It would require up-to-date record mainten-
ance the like of which does not exist now. And it would require
significant amounts of time and money. However, until information
of this nature is gathered and carefully studied, we really cannot
assess the effectiveness of the educational offerings of the Michi-
gan State University School of Police Administration and Public
Safety. we do not, at the present time, have much accurate informa-
tion on how college success relates to job success; nor do we know
how academic ability, when considered apart from academic success,
relates to job performance. And finally, apart from some scores
on academic tests, we do not have any accurate estimation of the
type of person who is attracted to the field of law enforcement.

In summary, there is a definite need for further study in the

VI-6

areas mentioned above and many reasons why the Michigan State Uni-
versity School of Police Administration and Public Safety should
assist in these efforts. Admittedly, the first attempt was not
particularly successful, but it can serve as a beginning. By
learning from earlier mistakes, and by utilizing available informa-
tion sources and consulting services, the Michigan State University
School of Police Administration and Public Safety can become a
leader in the study and analysis of educational offerings in the
various areas of law enforcement and how much these educational
offerings actually accomplish. As a result of these studies, the
School will be able to continually improve its course offerings,

outside services, and general value to the community.

B~1

BIBLIOGRAPHY

Books

Buras, Oscar R., The Fifth Mental Measurement Yearbook.

Cronback, Lee J., Essentials of Pa cholo ical Testing, 2nd ed.
New York: Harper and EFotEers, I965.

Freeman, Harold, Introduction to Statistical Inference. Reading,
Pa.z Wesley b is ing ., Inc., .

 

Hathaway, Starks R., and Monachesi, Elio 1)., Analyzing and Pro-

dicti Juvenile'Delin uenc ‘with the MMPI. Minneapolis:
CEIweII Piess, Inc., 1353.

Ostle, Bernard, Statistics in Research. Ames, Iowa: The Iowa
State College ess, 4.

Rand Corporation, A.Million Random Di its with 100 000 Normal
Deviates. Glencoe, IIl.: The Free PESss, I955.

 

Journal Articles

Aiken, Lewis R. Jr., "Some Nomographs for Academic Prediction

Work,” Educational and Po cholo ical Measurement, XXIV,
(Fall, TW—L—L—w 4 , 1 -9 . .

. ”The Prediction of Academic Success and Early Attrition
by Means of a Multiple-Choice Biographical Invente ,“
American Educational Research Journal, (March, 1964 , 127—
135.

Bachman, Gerald 0., ”Prediction of Academic Aehievement Using the“
Edwards Need Achievement Scale,” Journal of lied Ps chol-
2a, XLVII, (February, 1964), 16-1§'_—. ‘lIE—' 'i "

Berger, Ben, and Hall, Everette, ”Personality Patterns and Achieve-

ment in College," Educational and Po cholo ical Measurement,
xnv, (Spring, 1960—SSW, 4'3'."' I'.. ”“" '

Barnette, Leslie I. Jr., ”A Structured and Semi-Structured Achieve-
ment Test Applied to a College Sample,” Educational and Po -
chological Measurement, XXI, (Summer, 1961), 647-656.

B—2

Bernhardt, Harold E. Jr., ”Intraception Test Score and Psychiatry
Grade as.A Freshman and as A Sophomore Medical Student: .A
Validation Study of.A Subscale of the Edwards Personal Pref—

erence Schedule," Educational and Pa cholo ical Measurement,
xx, (Summer, 1960)"56', 5"‘-3'79"'.' "" ""“" i' '”""'"'""
Boe, Erling E., "The Prediction of Academic Performance of Engi-

neeri Students," Educational and Psychological Measurement,
XXIV, spring, 1964 ’ - O

Boyce, Richard It, “The Prediction of Achievement in College

Algebra," Educational and Psychological Measurement, XXIV,
(Spring, 1 1 " 0

Brown, Frederick G. and Iolins, Leroy, ”An Empirical Evaluation
of the American College Testing Program,” The Personnel and
Guidance Journal, (January, 1965), 451-456. "”

Cash, I. L. Jr., ”Predictive Efficiency of Freshman Entrance Tests,"
Journal of Psychological Studies, (June, 1962), 111-116.

."Relationship of Personality Traits and Scholastic Apti—
ude to Academic Achievement in Theological Studies," Journal
of Plzch01051031 Studies, (June, 1962), 105-110. _—

Chahbasi, Powis, "Use of Projective Tests in Predicting College
Achievement," Educational and PI cholo ical Measurement, 21,
('inter, 1960)_5'§9'T, - 4T. " ‘""' L._£._.._...._._._.

Demas, George D., and Spolyar, Ludwig J ., "Academic Achievement of
College Freshmen in Relation to Edwards Personal Preference

Schedule," Educational and Po cholo ical Measurement, XXI,
(Spring. 19FT'T—""L—“'———1 . 4 -479.

Tomino, George, ”Comparison of the D48, Cattell Culture Fair and
Army Beta Tests in A.Smmple of College Males," Journal of
Consulting Psychology, (October, 1964), 468-469.

Ellis, Kenneth E., "How Effective is Differential Prediction in ‘
Three Types of College curricula," Educational and Po chol-
ogical Measurement, XXI, (Spring, 1961), 459-471.

Endler, Norman 8. and Steinberg, Danny, ”Prediction of Academic

Achievement at the University Level,” Personal and Guidance
Journal, VIII, (1963), 694-699 M

French, John E., "Aptitude and Interest Score Patterns Related to
Satisfaction with College Major Field," Educational and Po -
cholo ical Measurement, XII, (Spring, 1961), 287-294.

B—3

Freelich, H. P. and Mayo, G. D., ”A.Note on Under- and Overaohieve—
ment Measure," Personnel and Guidance Journal, (March, 1963),

Gallagher, Joseph I. and Spencer, Richard E., ”Prediction of Success
in Basic German at the College Level,” Educational and PI
chological Measurement, XXIV, (Fell, 1964), 955-965.

Goodstein, Leonard D. and Reilbrun, Alfred B. Jr., ”Prediction of

College Achievement from the Edwards Personal Preference Sched-
ule at Three Levels of Intellectual Ability,” Journs; of

Applied Psychology, XLVI, (say, 1962), 317-320.

Griffen, Mary Louise, and Flaherty, Sister M; Rita, ”Correlation of
CPI Traits with Academic Achievement,~ Educational and Po -
cholo ical Measurement, XXIV, (Spring, I964), 5695372.

Hansmeier, Thomas I}, "The Iowa Tests of Educational Development as
Predictors of College Achievement," Educational and PI chol-
o ical Measurement, xx, (Iinter, 196W81T—‘L", 4 45.

Hewitt, John R. and Rosenberg, Leon A., "The MMPI as A Screening
Divice in an Academic Setting," Educational and Pa cholo ical
Measurement, XXII, (Winter, 1962), I29-I37.

Ivanoff, John M., Malloy, John P., and Base, Janet R., ”Achievement,
Aptitude and Biographical Measures as Predictors of Success

in Nursing Training,” Educational and PI cholo cal Measure-
ment, XXIV, (Spring, 1964), 389-39I.

Juola, Arvo E., ”The Development of an Academic Predictor Scale
Based on Students' Attitudes Toward Education,” The Personnel
and Guidance Journal, XLII, (December, 1963), 381- .

King, Paul, Norrell, Gwen, and Erlandson, P. L., ”The Prediction of
Academic Success in.A Police Administration Curriculum," lump

cational and Psychological Measurement, m, (Iinter, 1959')?
4 - 51o

Klugh, Henry E. and Bierley, Robert, “The High School and College
Ability Test and High School Grades as Predictors of College

Achievement,” Educational and Pa cholo ical Measurement, XIX,
(Iinter, 1959), 625-626.

Lewis, John E., "Correlates of Persistence in.A College of Pharmacy,“'

Educational and Peychological Measurement, XXIV, (Spring, 1964),
5- .

. ”Pro-College Variables as Predictors of Freshman, Sopho-

more and Junior Achievement,” Educational and PI cholo ical
Measurement, XXIV, (Spring, 1964), 353-556.

 

B-4

. ”The Relationship of Selected variables to Achievement-
and Persistence in A Masters Program in Business Administra-

tion,“ Educational and Ps cholo ical Measurement, XXIV,
(Minter—M l9 4 , 951- 5'4'.'l_—'E'__'_—_

Lewis, John It, Braskamp, Larry and Statler, Charles, ”Predicting
Achievement in.A College of Law,“ Educational and Po chol-
ogical Measurement, XXIV, (Fell, 1§325, §ZT-§Z§.

Mann, Sister M; Jacinta, S. C., “The Prediction of Achievement in
a Liberal Arts College,” Educational and Pr cholo ical
Measurement, XXI, (Spring, l§31’, 435-355.

Michael, lilliam R., et al., "A Note Concerning the Predictive

Validities of Selected cognitive and Hon-Cognitive Measures
for Freshman Students in A.Liberal Arts College,” Educa-

tional and Pe cholo ical Measurement, XXIV, (Spring, l§64),
373-375- ‘

Michael, Iilliam R., et al., "High School Record and College Board
Scores as Predictors of Success in.A Liberal Arts Program-

During the Freshman Year of College,” Educational and PI -
chological Measurement, XXII, (Spring, 1535’, §§§-455.

Miller, Adam, "General Ability and Interest Measure as Differenp
tial Predictors of Academic Achievement,“ Educational and

Pezchologioal Measurement, XXIV, (Spring, I§Bzi, 357-335.

Pippert, Ralph, and Archer, Sidney R., "A.Cemparison of The Meth-
ods for Classifying Uhderachievers with Respect to Selected
Criteria,“ The Personnel and Guidance Journal, (by, 1963),
788—891.

Robertson, Malcolm, and Hall, Everett, ”Predicting Success in

Graduate Study,“ The Journal of General Pe cholo , (Oct-
ober, 1964), 359-535.

Schults, Richard E., and Baker, Robert L., “A Hector Analysis of
the Kuder Preference Record — Occupational, Form D,“ Edu—

cational and Pa cholo ical Measurement, n11, (Spring—, 1962),
§TZIUIT'"”""JL""JL"""""""'

Sean, Matthew R., "An Alternative Approach to Scholastic Aptitude
Tests as Predictors of Graduation Rank at Selected Colleges,”

Educational and Psychological Measurement, XXIV, (Spring,
9 4 9 34 ’ 520

Sharp, Bert L., ”College Achievement: Its Relationship to High
School Achievement Experiences and Test Scores,” The Per-
sonnel and Guidance Journal, XLI, (Summer, 1962),'-T247- so.

3’5

Spaulding, Helen, “The Prediction of First Year Grade Averages in
A Private Junior College," Educational and Fe cholo cal
Measurement, XIX, (Winter, 1959,, 327-353.

Vick, Mary Catherine, "Predicting GPA at A ball Southern College,"

Educational and Pszchological Measurement, XXII (linter,
129795-79e

lerwick, Charles E., "Relationship of Scholastic Aspiration and
Group Cohesiveness to the Academic Achievement of Male Fresh-
men at Cornell university,” Human Relations, XVII, (no. 2),

wetley, Donivan J., "Type, Location, and Size of High School and
Prediction of Achievement in an Institute of Technology,“

Educational and Pezchological Measurement, XXIV, (Spring,
9 " e

letley, Donivan J. and Merwin, Jack 0., "The Effectiveness of Veri-
ables for Predicting Academic Achievement of Business Stu-
dents," The Journal ofggperimental Education, (Iinter, 1964),

OTHER SOURCES

Hathaway, Starks R., and McKinley, Charnley J., Booklet for the
Minnesota Multi hasic Personalit Invento , Rew York: The
Piychological Cgrporation.

Hickok, Benjamin J ., Eglish Placement Test, Form _A_, East Lansing:
Michigan State University.

Juola, Arvo E., comparative Standings of the Various College and
Curriculum Groups on the Orientation leek.Hkaminations, East
Lansing: Michigan State University, Fell Terms, 1959, 1960,
1961.

Michigan State university, Reading Test, Form A62, Prepared by The
Office of Evaluation Services, East Lansing: Michigan State
University.

Michigan State university, The Gee of Orientation Test Data, A
Report Prepared by the Office of Evaluation Services, East
Lansing: Michigan State University, 1960.

Otis, Arthur 8., Otis Quick-Scoring Mental Ability Tests, Gamma
Testj Form EM, Hes York: The lorld Book Company, 1954.

Peychological Corporation, College Qualification Tests, combined
Booklet Edition, Form C. Rew York: The Psychological Cor—
poration, 1959.

APPENDIX

POLICE ADMINISTRATION TESTING

NAME: DATE:

 

Orientation Test Scores:

CQT V I N T 8 E A R

 

 

 

 

Otis Mental Abilitz Test: IQ score___ Below Av.‘ Av.__ Above Av.—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

, Prof. Psxch.

Chemist

 

 

 

 

 

 

Civil Eng. Min.-Met. Eng.
moon. Eng. Druggist
Couns. Psych. Job Print.

 

 

 

 

 

£9233 Occup. £13; , 2 (verif. S__D___U__) pg;

Title D.S Title D.S. Trait Raw Score _
}Ag£i;_£gent Journ L

Elect. Eng Archit. F

Farmer Lagyer K

Forester Ret. Cloth. He

Minister ‘Ins. Agent D

Editor Dentist EV

Physician vetern. Pd

lClin. Psych. Indust. Eng. Mf

”Ind. Psych. Pediatro Pa

HY.M.C.A. Sect. Psychiatr. Pt

School Sup. Rad. thr. Sc

‘Agct. Int. Decor. Ma

meteor. H.S. Counsel Prob. Areab____
Pers. thr. H.S. Sci. Tea. No Prob. Area;____
Sales H.S. Math. Tea

Strong vocational Interest Test

c a rker
nister
musician
C
nior
coun
ca

81‘

c s
as
. as
Life Insur. Sales
r.
or
thor-Journ.
Pres ncern

 

"I7'1?@“E'EIJ‘IEMMWQIEEFS