GENDER - RELATED EFFECTS OF ADVANCED PLACEMENT COMPUTER SCIENCE COURSES ON SELF - EFFICACY, BELONGINGNESS, AND PERSISTENCE By Jonathon Andrew Good A DISSERTATION Submitted to Michigan State University i n partial fulfillment of the requ irements f or the degree of Educational Psychology and Educational Technology Doctor of Philosophy 2018 ABSTRACT GENDER - RELATED EFFECTS OF ADVANCED PLACEMENT COMPUTER SCIENCE COURSES ON SELF - EFFICACY, BELONGINGNESS, AND PERSISTENCE By Jonathon Andr ew Good The underrepresentation of women in computer science has been a concern of educators for multiple decades. The low representation of women in the computer science is a pattern from K - 12 schools through the university level and profession. One of the purposes of the introduction of the Advanced Placement Computer Science Principles (APCS - P) course in 2016 was to help broaden participation in computer science at the high school level. The design of APCS - P allowed teachers to present computer scien ce from a broad perspective, allowing students to pursue problems of personal significance, and allowing for computing projects to take a variety of forms. The nationwide enrollment statistics for Advanced Placement Computer Science Principles in 2017 had a higher proportion of female students (30.7%) than Advanced Placement Computer Science A (23.6%) courses. However, it is unknown to what degree enrollment in Th self - efficacy, belongingness, and expected persistence in computer science. A nation wide sample of 263 students from 10 APCS - P and 10 APCS - A courses participated in the study. Students completed pre and post surveys at the beginning and end of their Fall 2017 semester regarding their computing self - efficacy, belongingness, and plans to c ontinue in computer science studies. Using hierarchical linear modeling analysis due to the nested nature of the data within class sections, the researcher found that the APCS course type was not predictive of self - efficacy, belongingness, or expectations to persist in computer science. The results suggested - efficacy declined over the course of the study. However, gender was not predictive of belongingness or expectations to persist in computer science. Students were found to have entered into both courses with high a sense of self - efficacy, belongingness, and expectation to persist in computer science. The results from this suggests that students enrolled in both Advanced Placement Computer Science courses are already likely t o pursue computer science. I also found that the science. This suggests that educators should look beyond AP courses as a method of exposing students to comput er science, possibly through efforts such as computational thinking and cross - curricular uses of computer science concepts and practices. Educators and administrators should also continue to examine whether there are structural biases in how students are directed to computer science courses. As for the drop in self - efficacy related to gender, this in alignment experiences in the course to not negatively influence their self - efficacy. Further research should designation and curriculum of APCS - A or APCS - P alone may not capture the myriad of ways in which teachers may be addressi ng gender inequity in their classrooms. Research can also examine how student interest in computer science is affected at an earlier age, as the APCS courses may be reaching students after they have already formed their opinions about computer science as a field. Copyright by JONATHON GOOD 2018 v This dissertation is dedicated to Maria, Greyson, and Genny the center of everything I love. vi ACKNOWLEDGEMENTS I would like to thank Aman Yadav for the many opportunities to wo rk at home and abroad, the patience while waiting for me to see the errors of my ways, and the moments of kindness toward my family. Jack Smith, for introducing me to Dewey and for the multiple years of writing guidance. Jen Schmidt, for going above and be yond in being a resource for analysis and interpretation, and for the having the patience that role required. Niral Shah, for reminding me to consider how the work we do affects students and teachers. Punya Mishra , for encouraging me to begin my journey, pushing me to embrace the half - formed idea, and all the multitude of ways in which you continue to shepherd me along the way. John Bell, for the opportunities to develop my skills and your humane guidance. CSEER group members past and present, for the shar ed laughter, triumphs, and food. Deep Play group members and collaborators, for helping me get my first work out the door, sharing stories, and occasionally keeping our meetings on topic. STEM+C colleagues, for welcoming a different type of researcher, and the opportunity to engage in unfamiliar and exciting methods of teaching. NASA Office of Education and Logistics and Technical Information Division for the unique opportunities to broaden my experiences this past year. My students past and present, for t he privilege of learning from our time together. Rohit, for all the fun, difficult, ridiculous, and serious moments from the first day of graduate school and onward. Swati, for pushing me to think beyond the dominant narrative. Danah Henriksen for mentorin g me in the ways of Punya , teaching, and dragons . Sarah G retter, for guidance on the dissertation process, writing, and key insights into the workings of Aman . Alex Lishinski, for endless statistical help and occasional outdoor adventures . Patrick Beymer, for statistical advice and letting me sleep on a futon instead of the floor . Adams , Teeters - Whitlock , and Crosby vii families , for the stress relief, childcare, car repairs, last minute assistance, musical interludes, and help with the GREs. Jill Metzger, for helping us limp across the finish line . Rich, Ray, and Tim, for the comic relief . Dad , for encouraging a dry sense of humor and buying the TI 99/4 so long ago . Mom , for encouraging me to take bigger risks in life and convincing me to stay in college 25 yea rs ago . Genny , for saying things out loud that the rest of us needed to hear, and the reminders to stay on task . Greyson , for long talks when we both should have been sleeping, and the significant assistance with gender research . Maria , for your unwavering love, supporting our family throughout this adventure, sharing the setbacks and victories, taking a big leap of faith with me, and reminding me of the only things that matter. viii TABLE OF CONTENTS LIST OF TABLES...................................... .....................................................................................x CHAPTER 1 : INTRODUCTION ................................ ................................ ................................ .. 1 1.1 Background ................................ ................................ ................................ ...................... 1 1.2 Theoretical Framework ................................ ................................ ................................ .... 4 1.2.1 Persistence of Female Students in Computer Science ................................ .............. 4 1.3 Belongingness ................................ ................................ ................................ .................. 4 1.4 Self - Efficacy ................................ ................................ ................................ ..................... 5 1.5 Positionality Statement ................................ ................................ ................................ ..... 6 1.6 Purpose Statement ................................ ................................ ................................ ............ 8 CHAPTER 2 : LITERATURE REVIEW ................................ ................................ ....................... 9 2.1 Belongingness ................................ ................................ ................................ .................. 9 2.1.1 Belongingness and Stereotype Threat. ................................ ................................ .... 11 2.2 Self - Efficacy ................................ ................................ ................................ ................... 13 2.3 Broadening Partic ipation in Computer Science ................................ ............................. 19 2.3.1 Timing: Reach Students Before University ................................ ............................ 19 2.4 Structural Barriers for Equity ................................ ................................ ......................... 22 2.4.1 Representations of Computing Disciplines ................................ ............................. 24 2.5 Context of the Study ................................ ................................ ................................ ....... 27 2.5.1 Advanced Placement Computer Science ................................ ................................ 27 2.5.1.1 Advanced Placement Computer Science A ................................ ..................... 28 2.5.1.2 Advanced Placement Co mputer Science Principles ................................ ........ 29 2.6 Research Purpose ................................ ................................ ................................ ........... 32 CHAPTER 3 : METHODS ................................ ................................ ................................ ........... 33 3.1 Sampling ................................ ................................ ................................ ......................... 33 3.2 Participants ................................ ................................ ................................ ..................... 36 3.3 Measures ................................ ................................ ................................ ......................... 37 3.3.1 Student Surveys ................................ ................................ ................................ ...... 37 Belongingness ................................ ................................ ................................ ................... 38 Self - efficacy ................................ ................................ ................................ ...................... 3 9 Pers istence ................................ ................................ ................................ ......................... 39 Prior Grades ................................ ................................ ................................ ...................... 40 3.3.2 Teacher Surveys ................................ ................................ ................................ ...... 40 3.4 Procedur es ................................ ................................ ................................ ...................... 40 3.5 Analysis ................................ ................................ ................................ .......................... 42 CHAPTER 4 : RESULTS ................................ ................................ ................................ ............. 44 4.1 Descriptive St atistics ................................ ................................ ................................ ...... 44 4.2 Outcome Variable Change ................................ ................................ ............................. 47 4.3 Normality of Outcome Variables ................................ ................................ ................... 50 4.4 Analysis and Results ................................ ................................ ................................ ...... 51 ix 4.4.1 Self - Efficacy Analysis and Results ................................ ................................ ......... 51 4.4.2 Belongingness Analysis and Result s ................................ ................................ ...... 54 4.4.3 Persistence Analysis and Results ................................ ................................ ............ 57 CHAPTER 5 : DISCUSSION ................................ ................................ ................................ ....... 61 5.1 Implications ................................ ................................ ................................ .................... 65 5.1.1 Implications for Practice ................................ ................................ ......................... 66 5.1.2 Implications for Future Research ................................ ................................ ............ 68 5.2 Limitations ................................ ................................ ................................ ..................... 70 5.3 Conclusion ................................ ................................ ................................ ...................... 73 APPENDICES ................................ ................................ ................................ .............................. 75 APPENDIX A Survey for Students ................................ ................................ ......................... 76 APPENDIX B Survey for Computer Science Teachers ................................ .......................... 83 REFERENCE S ................................ ................................ ................................ ............................. 86 x LIST OF TABLES Table 1: Student Participant Summary by Class ................................ ................................ ........... 35 Table 2: Student Participant S ummary by Gender and Grade Level ................................ ............ 37 Table 3: Student Participant Summary by Race and Course ................................ ........................ 37 Table 4: Cronbach's Alpha for Bel ongingness Scale ................................ ................................ .... 39 Table 5: Descriptive S tatistics for Self - Efficacy, Belongingness, Persistence, and Student Prior Grades ................................ ................................ ................................ ................................ ........... 44 T able 6: Self - Efficacy, Belongingness, Persistence, and Student Prior Grades Correlations ...... 45 Table 7: Frequency of Student Participants by Course and Gender ................................ ............. 46 Table 8: Frequency of Teacher Gender by Course ................................ ................................ ....... 46 Table 9: Frequency of Teacher Gender for Each Class Section by Course Type ......................... 47 Table 10: Descriptive Statistics for Self - Efficacy by Student Gender, Course, and Teacher Gender ................................ ................................ ................................ ................................ ........... 48 Table 11: Summary Statistics for Belongingness by Student Gen der, Course, and Teacher Gender ................................ ................................ ................................ ................................ ........... 49 Table 12: Summary Statistics for Persistence by Student Gender, Course, and Teacher Gender 50 Table 13: Outcome Variable Skew and Kurtosis ................................ ................................ .......... 51 Table 14: Self - Efficacy Hierarchical Linear Model Analysis ................................ ...................... 54 Table 15: Belong ingness Hierarchical Linear Model Analysis ................................ .................... 57 Table 16: Persistence Hierarchical Linear Model Analysis ................................ .......................... 60 1 CHAPTER 1 : INTRODUCTION 1.1 Background Within the United States of America and across the globe, computer science education is enjoying a renaissance in interest from both the general public and policymakers (Code.org, 2017; Obama, 2016; Partovi, 2013; Trump, 2017) . Th is resurgence in interest in computer science education is bolstered by a combination of factors, such as the increasing focus on STEM (Science Technology Engineering Mathematics) education, a shortage of programmers, and computer science being perceived a s a path for economic advancement (Google Inc. & Gall up Inc., 2015) . While the overall enrollment numbers in CS are increasing, the share of female students earning bachelor degrees in CS has fallen from a high of 37% in 1984, to 17.5% in 2015 (U.S. Department of Education, 2017) roughly half of scie nce and engineering degrees as defined by the National Science Board (2016) . Within this range of science and math ematics majors , though, a few majo rs continue to produce a low proportion of female graduates: computer science, engineering, mathematics, and statistics. The number of female students participating in computer science at the K - 12 level also remains abysmal. In 2016, female students in the United States only accounted for 23.2% of students taking the high - school course Advanced Placement Computer Science A (APCSA) exam (The College Board, 2016b) . In Mississippi and Montana, no female students took the APCSA exam in 2016. E nrollment of women in APCSA is abysmal when compared to another AP course such as AP Calculus AB, which has comparable demands and preparatory courses; female students comprised 49.4% of students taking the AP Calculus AB exam in 2016 (Ericson, 2014; The College Board, 2016b) . These trends of low participation of women in computer 2 science reverberates in the workforce, where the culture of computing does not always provide women with a sense of be longingness (Aspray, 2016; DuBow, Kaminsky, & Weidler - Lewis, 2017; Margolis, 2013; Margolis & Fisher, 2003) . DuBow, Kaminsky, and Weid ler - Lewis (2017) found that women find that receiving respect, e ncouragement, and support from classmates and colleagues is a significant factor in helping them persist within the computing field. All students need to be provided equal opportunity and support to enter and persist in computer science. On a macro - econom ic level, the need to have gender equity is highlighted by the fact that countries and industries with a gender imbalance in the workplace suffer an economic cost to their efficiency and overall production (Dollar & Gatti, 1999; Plantenga, 2015) . On a micro level, computer scientists and software developers bring their own experi ences and perspectives to the design of software and hardware, often testing the products themselves and trying to anticipate what a user needs or will do. A largely male workforce often develops products that appeal to male consumers and continues to per petuate the male - dominated environment (Crowell, 2016; Oudshoorn, Rommes, & Stienstra, 2004; Rommes, Oost, & Oudshoorn, 1999) . Oushoorn, Rommes, & Stienstra (2004) provided a case stud y of an online communit y , which used primarily male designers, and as a result the hardware and software design choices led to further gender bias in participation. For these reasons, and for social justic e concerns, we should be working to increase the proportion of women in CS programs. One way to do this is to engage female students in computing within K - 12 classrooms with the intent to both recruit and retain female students that currently are not persi sting in computer science fields. As scholars have tried to address the problem of gender equity within computer science education, they have pursued multiple approaches. 3 One approach that educators have taken to broaden the appeal of computer science t o female students is to present how computer science plays a role in multiple disciplines, focus on non - computing application of computer science principles, and encourag e pedagogical strategies shown to support gender equity (Goode, 2008; Kafai & Burke, 2014; Margolis & Fisher, 2003; Ryoo, Goode, & Margolis, 2015; Yadav, Gretter, & Good, 2017) . In an exampl e of one such effort to increase the number of female students in computer science , College Board recently introduced the Advanced Placement Computer Science Principles (APCS - P) course that portrays computer science as a broader discipline. College Board d eveloped the APCS - P in an effort to (The College Board, 2016a, p. 4) . In contrast with APCSA, which is largely a traditional programming course with programming lab assignments and assessments in a single language, the APCSP framework allows teachers to decide on which language to use for the course, increase in the percentage of female students in APCSP (30.7% female, 69.3% male) when compared to APCSA (23.6% female, 76.4% male) (College Board, 2017; The College Board, 2016b) , this is only the first step in increasing gender equity within computer science. Until future enrollment and attrition rates in undergraduate CS programs become available, we will not have a direct measure of whether enrollment in the APCSP course will result in sustained interest in computer sci ence at the college level and beyond. What can be done in the interim to gain insight into whether these curricular efforts are likely to show promise in addressing the gender equity issues within the computer science field? 4 1.2 Theoretical Framework 1.2.1 Persiste nce of Female Students in Computer Science While recruiting female students into computer science remains a major area of focus, we also need to examine how their experience in CS might influence their persistence . Even if efforts to increase female stude male students, their experiences have to be meaningful and rewarding for female students to want to persist (Google Inc. & Gallup Inc., 2014; K afai & Burke, 2014; Margolis & Fisher, 2003) . Historically attrition of female undergraduates from CS is higher than that for male students and a significant factor in the lower number of degrees awarded to women (Chen, 2013; Cohoon & Lord, 2006; Hamilton et al., 2016) . Prior work in other academic subjects has suggested that belongingness and self - efficacy are significant predictors of persistence (Beyer, 2014; Good, Rattan, & Dweck, 2012; Lent, Brown, & Larkin, 1986) . 1.3 Belongingness Baumeister and Leary (1995) drive to form and maintain at least a minimum quantity of lasting, positive, and significant Deci and Ryan (2000a) , within the framework of self - determination theory, described belo ngingness as a component psychological needs to guide them toward more competent and socially integrated behavior. Good, Rattan , and Dweck (2012) how it evolved over time, and affected persistence in the discipline. They foun d that while male was eroded when exposed to an environment that reinforces negative stereotypes about female 5 mathematicians (e.g. women are not good mathematicians ) or the belief math ematic ability is a fixed trait. Similarly, Lewis et al. (2016) of belongingness in physics was more impactful for women with STEM careers than men. If a student continues to encounter situations that reinforce existing negative stereotypes, the probability of experiencing student has a reduced sense of belonging and likelihood to pursue further studies in the field (Cun diff, Vescio, Loken, & Lo, 2013; Steele, 1997) . Within computer science education, Cheryan et al. (2009) found that removing stereotypically male - identified objects from a computer science classroom such as Star Trek posters and video games resulted in a high sense of belongingness for female students w hile having little to no effect on male students. While there has been some research on how CS their sense of belongingness influences their self - efficacy in CS as well as their persistence in CS. 1.4 Self - Efficacy Bandura (1997) described self - iderable research has been done on the relationship between self - efficacy and student learning outcomes and academic performance (Bandura, 1977, 1986, 2002; Lent, Brown, & Hackett, 1994; Lent et al., 1986) . Research has also suggested that self - and domain self - efficacy is predictive of career choice (Bandura, 2002) . Lent and colleagues (1994, 1986; 2002) cognitive career framework and argued that domain self - efficacy was a main predictor of career 6 choice and persistence across multiple fields of work. Pajares and Valiante (1997) found that writing self - efficacy contributed to fifth - mediating role for existing writing ability, and thus suggested that teachers should pay close att these are predictive of future academic choices. Within computer science and related fields, the same relationship can often be found in the literature with se lf - efficacy being tied to higher performance, persistence, choice of academic majors, and career aspirations (Beyer, 2014; Blaney & Stout, 2017; Cohoon & Lord, 2006; Lishinski, Yadav, Good, & Enbody, 2016) . Thus, self - efficacy becomes a useful indicator for predicting not only academic performance, but also of the students persistence in the field. Prior work in computer science education has examined the role of physical classroom space as well as participation in introducto belongingness and self - efficacy (Beyer, 2014; Cheryan et al., 2009; Lishinski et al., 2016) . However, there is limited research on how presenting computers science as being applicable to larger problem solving, across multiple disciplines, has an effect of self - efficacy and belongingness , especially at the K - 12 level. 1.5 Positionality Statement I am a researcher of computer science education with a focus on equity, hav ing formerly worked as a teacher and technology coordinator in PreK - 12 settings. My personal and professional background both limits and informs my work in computer science education and provides a lens I bring to equity issues . I worked at two private and independent schools over the span of 12 years in Virginia and Ohio. As a result, I worked with students with financial means , with some students from middle and lower income families , who were provided financial aid to 7 be able to attend the school . The school in Virginia was a rural boarding school, grades 8 - 12, with both boarding and daily commuting students. The technology classes at this school often had all male students, or extremely low (e.g. , 15:1) ratios of male to female students. I also found the classes to be overwhelming populated by white males and male international students from eastern Asia. The other school I taught at in Ohio was an all - female day school, located in an affluent suburb. B oth schools led me to question the gender and racial composition of the schools, along with how gender and racial imbalance appeared exacerbated in my computer science and technology classes. Being a white male, I bring my own biases in what I noticed and did within a classroom. Particularly at the school in Ohio, which to its credit had a strong focus on issues of social justice, I noticed how the contrast in gender balance between the two schools seemed to influence how students engaged with the subjects. I had also now had two children of my own and became increasingly aware of sexist assumptions in my own daily life and work that troubled me. I adjusted my teaching methods to attempt to address some of these issues, but found the recommendations given to me sometimes conflicting or confusing. T he engagement I saw in my female students encouraged me, particularly because it contrasted with the narratives I was hearing from other schools regarding female students and technology. The students were excited to engage with the difficult technical pro blems of robotics and programming, appeared to be comfortable displaying technical proficiency, and openly rejected the idea of technology being a domain they were discouraged from entering. Having the background and experiences as a teacher informed my se lection of research topics in computational thinking, computer science education, and related efforts in gender equity in K - 12. I recognize the limitations of my own experiences, particularly being male and 8 white, and how those limitations influence condu cting research that includes perspectives beyond my own. I carry the preconceptions that come from my background, social status, and privilege in these systems. Similar to my examining of my own teaching while in the K - 12 environment, I struggle to exami ne how my research methods may have overlook ed other perspectives . My hope is that my research, through these acknowledgements and resulting adjustments, will provide some insight for researchers, teachers, and students of varied backgrounds to address e quity issues in computer science. 1.6 Purpose Statement The purpose of this study was to examine how enrollment in two different Advanced Placement Computer Science (APCSA vs APCS - of belongingness, self - efficacy , and persistence within computer science, and how student gender may interact with that relationship. These variables of interest were chosen for their relationship to the likelihood of a student choosing to pursue computer science in the future, includin g at the undergraduate level as well as persist within the field. 9 CHAPTER 2 : LITERATURE REVIEW While recruiting females into computer science remains a major area of focus in the eriences influence their persistence in CS . If we want to increase female student enrollment in computer science to parity with male students, we need to provide meaningful and rewarding experiences for female students to want to persist (Good et al., 2012; Google Inc. & Gallup Inc., 2014; Kafai & Burke, 2014; Margolis & Fisher, 2003) . Historically, attrition of female undergraduates from CS is higher than that for male students and a significant factor in the lower number of degrees a warded to women (Chen, 2013; Cohoon & Lord, 2006; Hamilton et al., 2016) . Prior work has suggested that belongingness and self - efficacy are significant predictors of persistence and academic achievement, in general (Goodenow, 1993; Hausmann, Schofield, & Woods, 2007; Pittman & Richmond, 2007) as well as in computer science (Lishinski et al., 20 16) . Understanding what influences both belongingness and self - efficacy in computer science classrooms, and how computer science educators may create an environment that bolsters in computer science. 2.1 Belongingness Baumeister and Leary (1995) stated that the need for belonging has two criteria: frequent, pleasant int Deci and Ryan (2000b) presented the need for belongingness as a motivational basis for learning, with belongingness aiding in the transfer of group knowledge to the individual. Picket, Gardner, and Knowles (2004 ) found that belongingness developed through verbal and visual social cues, and could affect basic cognitive functions. Good, Rattan and Dweck (2012) examined belongingness in the context of mathematics education, how it evolved over time, and affected student 10 persistence in the discipline. The authors used a meas ure of belonging with five factors of belongingness while enrolled in a calculus course. They found while both male and female ematics were significantly predicted by their sense of reinforces negative stereotypes about female mathematicians (e.g. statements that women are not good mathematici ans) or the belief that math ematic ability is a fixed trait. Lewis et al. (2016) provided practical recommendations for a ddressing gender equity for physics educators based within the fi students that feelings of not belonging in the domain are normal at first and fade over time, (4) use cooperative pedagogical strategies, such as a jigsaw activity model, to encourage meaningful social interaction within the classrooms, and (5) and to tie the course content to a larger social context outside of the classroom to help affirm the value of what is being learned. Within computer science education, Cheryan et al. (2009) conducted a series of four studies exa and their classroom environments. In the first study, the authors developed a list of objects likely to be found in the office of a stereotypical computer scientist from two separate groups of undergraduate students. The researchers then decorated three classrooms, one with stereotypical objects (e.g. Star Trek posters), one with non - stereotypical object (e.g. nature posters), and one without decorative objects, where they con ducted surveys of undergraduates about their likelihood to major in computer science. The researchers found that while male student interest 11 in majoring in computer science did not vary across rooms, female students in the non - stereotypical room had a hig her level of interest in majoring in computer science than females in the stereotypical and bare room. In their second study, the researchers described to undergraduate women two potential workplaces, both with similar salaries and an all - female team of co workers, that differed only in the types of objects (stereotypical and non - stereotypical) found in their workplace. The students reported a lower sense of belonging in the stereotypical office than the non - stereotypical office, in spite of the presence of an all - female team. In the third study, undergraduate students of both genders were given descriptions of gender - balanced workplaces with similar salaries, one described using the stereotypical objects and one without. Male students were more likely than female students to choose the stereotypical workplace, however, overall males and females preferred the non - stereotypical workplace. Finally, the fourth study used a similar design to the third study, but the workplace was described as a web development company. The authors found that male students preferred the stereotypical environment over the non - stereotypical, while female students preferred the non - stereotypical environment. These findings point to not only the importance of environment in determi ning gender composition of the workplace, and how certain computing disciplines (e.g. web design) can convey gendered messages for students. 2.1.1 Belongingness and Stere otype Threat. Belongingness is often intertwined with concerns about stereotype threat (Steele, 1997) in that both involve students being aware of how aspects of their own identity, such as race or gender, relate to their own conceptions of who does or does not become a me mber of a given field/domain. Steele described negative stereotypes as typecasting a group of people, such as 12 nt] that the existence of such a stereotype means that anything one does or any of one's features that conform to it make the stereotype more plausible as a self - characterization in the eyes of others, and perhaps even in ng this predicament, while also completing the normal tasks required in that domain, such as completing homework or preparing for tests, can have a performance is due to the mental energy spent upon trying to avoid being perceived as an example of a negative stereotype (Murphy, Steele, & Gross, 2007; Steele, 1997; Steele & Aronson, 1995) . Stereotype threat appears to only negatively affect members of groups with a negative stereotype . W woman or person of color might (Steele, 1997) . Inzlicht and Ben - Zeev (2000) found that female undergraduate students assigned to both single gender and mixed gender three - person work groups performed better in math ematic s and verbal assessments in all - female work groups. Inzlicht and Good (2005) gender, race, and ethnicity can affect the strength of stereotype threat, with its effects being more pronounced when students perceive themselves to be outnumbered in the classroom. This is further exacerbated when there is a mismatch between the teacher and student demographics in the classroom (Marx & Roman, 2002) . The teacher may be perceived as another indicator of who works in a particular f ield or discipline, and thus can either reinforce or help counteract a negative stereotype. As a student continues to encounter situations that reinforce existing ich 13 the student has a reduced sense of belonging and likelihood to pursue further studies in the field (Cundiff et al., 2013; St eele, 1997) . Smith et al. (2015) similarly found that undergraduate female physics students identified greater stereotype threat than biology students, resulting in lower sense of identif ication with the field, and likelihood to continue in the field. Eccles et al. (1999) also showed that choice of major and courses at the university level can be affected for members of a group described by a negative stereotype. Tellhed et al. (2017) found that female correlated with lower STEM self - efficacy and sense of belongingness. Meanwhil e, male feminine, was also predictive of lower interest in entering those fields. While there has been some research on how CS classroom environment can influence s limited research on how their sense of belongingness influences their self - efficacy in CS as well as their persistence in CS. 2.2 Self - Efficacy Bandura (1997) described self - efficacy as there were several sources of self - efficacy including mastery experience, vicarious experience, persuasion, and a rousal. Mastery experience can be thought of as the successful completion of a task or challenge, leading to a heightened sense of self - efficacy in that particular task. For example, i n a computer science context, this may be that a student successfully compiles their code without errors for the first time, and the positive experience with that task helps bolster their self - efficacy. Inversely, they could spend a significant amount of time searching for a syntax error that is preventing code 14 from compili ng, only to give up out of frustration. This could lead to a lessened sense of self - efficacy as a programmer and deter them from continuing their efforts. Vicarious experience, learning from observing someone else complete the task, could also increase For example, s tudents may experience this by watching peers solve coding problems that they were unable to do independently. Using the syntax error example from above, a student may seek out assistance, watch how their peer uses a method to find the syntax error and as a result feel that they could also use this technique to solve the issue. Likewise, if they seek out help an d sion, this may reinforce the negative effect on their self - efficacy in programming. Persuasion is the act of other persons encouraging you to complete a task, in a constructive manner, that bolsters self - efficacy to complete the task. For example, t eacher s may encourage a student to persevere in find ing the syntax error that is frustrating them or congratulate them on successfully completing a difficult task. This persuasion may also come from peers that try to convince them that this is a difficult task, their level of effort is normal, and that they should keep working on the problem . Self - efficacy is again bolstered if the student receives these messages to persevere. In a negative example, a teacher or student could send a message that the tasks is e xceedingly easy and that they should have finished earlier, and thus add ing - efficacy. Finally, arousal is largely concerned with any physio - emotional state, such as nervousness or confidence, that results in cha - efficacy regarding their ability to complete the task. For example, i f a student experienc es stress and frustration, it may be interpreted by the student as a sign of their lack of skill or ability , which could lead to 15 low er self - efficacy. However, if a student is in a more positive emotional state, such as in a classroom that is devoid of distraction or negative imagery, they are more likely to have higher self - efficacy. Of these four influences on self - efficacy, mastery experiences are often believed to be the most powerful (Bandura, 1977, 1986, 1997) as they are most authentical ly experienced by the subject. Considerable research has been done on the relationship between student self - efficacy and their academic performance (Bandura, 1977, 1986, 2002, Lent et al., 1994, 1986) . (1977, 1986) work on social cognitive theory suggested that self - efficacy was a strong predictor of performance on tasks of varying difficul ty. Bandura (1977) described how adults w ith a fear of snakes watched a boa constrictor being handled by an assistant in various manners, while building up to handling the snake themselves. By surveying the participants self - efficacy beliefs regarding the handling of snakes throughout the proce ss, he found that the direct experiences where participants handled the snake themselves were more powerful in increasing their self - efficacy than watching the assistant handle the snakes. Research has also suggested that self - efficacy in a particular doma in is predictive of career choice (Bandura, 2002) . Chemers and colleagues (2011) surveyed 665 undergraduates, graduate students, post - doctoral fellows, and recent graduates associated with a professional science organiza tion for C hicano and N ative A merican students. The researchers were investigating how personal psychological traits (e.g. self - efficacy , personal identity) mediated the influence of science esults from the path analysis suggested that the effects of research experience, community involvement, and mentoring on a efficacy, and identity as a scientist. Th e mediating effects were present in both undergraduate 16 and recent graduate, yet even stronger in the graduate and postdoctoral participants. This points to a need for instructors to understand - efficacy in order to ens ure that support experiences (e.g. mentoring) are more effective in improving the retention of students within their field. Pajares and Miller (19 94) surveyed 350 undergraduate students regarding their math ematic self - efficacy, perceived usefulness of mathematics, mathematic anxiety, mathematic self - concept, and prior experience. Immediately after completing the survey, students then were asked to complete a mathematic problem instrument as a measure of performance. The researchers found that mathematic self - efficacy had the strongest effect on mathematic problem solving and also mediated the effect of gender and prior experience on mathematic sel f - concept, perceived usefulness of mathematics, and mathematic problem - solving performance. This points to self - efficacy as one possible influence to counteract any negative performance concerns related to student gender and experience. In another study, P ajares and Valiante (1997) - efficacy as it related to their writing performance. Students completed a survey regarding their self - efficacy, perceived usefulness of writing, and writing apprehension in a single class period and then in another session completed a 30 - minute essay wri ting task. Teachers were also asked to rate the essay task. The researchers found that writing self - efficacy contributed significantly to fifth - grade student - efficacy also partially mediated the effects of existing writing aptitude and student gender on studen apprehension, perceived usefulness of writing, and writing performance. These findings 17 emphasize the significant role self - pay close attention to not only mediate the predictive power of past performance on future performance. Within computer science and related fields, similar relationship can be found with self - efficacy being tied to higher performance, persistence, choice of academic majors, and career aspirations (Beyer, 2014; Blaney & Stout, 2017; Cohoon & Lord, 2006; Lishinski et al., 2016) . For example, Beyer (2014) surveyed 1319 first - year students of all majors across three ye ars regarding their demographic information, stereotypes regarding CS, computer self - efficacy and experience, personality variables, and experience in CS courses. Student enrollment and grades in CS courses were tracked for their first year for 128 studen ts, who enrolled in a CS course. The researchers found that first - year undergraduate male students had higher levels of computing self - efficacy than female students, female students rated their own CS ability lower than their male counterparts, and that female students were more likely than male students to believe that women have as much ability as men in CS. As for predicting the likelihood to enroll in a CS course, the strongest predictors were computer self - efficacy, high interest in CS, low family or ientation, low openness to experiences, and low conscientiousness. Again, boosting computing self - efficacy could offer a manner in which to address disparities in whether a female student is likely to enroll in computer science courses. Blaney and Stout (2017) collected data from 2184 undergraduate students enrolled in introductory computing courses a cross 65 universities regarding their computing self - efficacy, sense of belongingness in computing, and perceived instructor inclusivity. The researchers were primarily interested in how self - efficacy and belongingness may differ in these courses between 18 first - generation college students and their peers, and how these differences appeared across gender. S - efficacy ha d a positive correlation with their level of interaction with faculty in class and student perception of inclusivity. Female students reported less interaction with their instructors both inside and outside of class. The implications the researchers offer ed is that instructors must make an effort to provide interaction (e.g. group discussion) within their courses for all students, encourage students to attend office hours to introduce themselves, and use explicitly inclusive language in the classroom (e.g. she/he, him/her) as they describe working in computing. Lishinksi et al. (2016) collected data from 346 undergraduate students enrolled in a CS1 - efficacy to their programming performance, and how the relationship changed over the span of the semester. The authors found that while male and female students performed similarly in the course, female students adjusted their self - efficacy beliefs to more accurately match their performance earlier in the course than male students, possibly internalizing early failures, which further lowered their self - efficacy beliefs. Male students were slower to adjust their self - efficacy beliefs to match their performance, often overestimating their abilities befor e eventually reaching a higher correlation between self - efficacy and performance later in the course. The authors argued that differences in the way that male and female students adjust their self - efficacy beliefs can be especially impactful because self - e fficacy beliefs can form a feedback loop with performance, where performance impacts self - efficacy, which further impacts future performance. One possible solution to this may be that instructors can examine their pedagogical choices to ensure students ar e not initially facing tasks that are too difficult or they run the risk of disproportionately lowering self - efficacy of female students over their male counterparts. 19 Considering the existing findings above (Beyer, 2014; Blaney & Stout, 2017; Cohoon & Lord, 200 6; Lishinski et al., 2016) , measuring self - efficacy becomes a useful measure for predicting not only academic performance, but point to the likelihood of a student continuing in their field. 2.3 Broadening Participation in Computer Science Interest and rese arch in gender equity within computer science over multiple decades has examined when we can best intervene in the educational pipeline, how we can remain aware of and address structural inequities in the existing school systems, and how we present compute r science as a discipline to students. Given the importance of belongingness and self - efficacy in existing curriculum and classroom practice to better address gender i nequality within computer science at the K - 12 level. 2.3.1 Timing: Reach Students Before University Prior work on increasing gender diversity in STEM fields has suggested that we need to engage women in high school and earlier (Google Inc. & Gallup Inc., 2014; Shapiro et al., 2015) . Shapiro et al. (2015) surveyed 1189 (414 male, 775 female) middle school students to examine op tions. Four hundred and seventy - five female students identified as girl scout members and 299 identified as not being girl scouts. The authors found that Girl Scouts had higher exposure to STEM career options than the non - Girl Scouts, along with being more likely to voice STEM career aspirations. The authors found that gendered notions of career paths had already been - Girl Scouts) anticipating taking a break from their ca reer to care for children, and boys being more likely to 20 state that men are better at some professions than women. The researchers pointed to the higher exposure to STEM experiences as Girl Scouts as possibly counteracting the negative effect of gendered c Google & Gallup (2014) surveyed 1600 part icipant s (600 male, 1000 female) that included pre - college students, college students, and recent graduates. The respondents were 50% pre - college and 50% were currently attending or had recently graduated from college. Fifty percent of the all respondent s were interested in or currently studying computer science or a related subject, while the remainder voiced no such interest. Results suggested that four factors rs - social encouragement, self - perception, academic exposure, and career perception - had largely been determined before female students entered the university and were less malleable after high school. Social encouragement was the strongest predictor of the decision to major in CS, accounting for 28.1% of the variance in the explainable factors. In addition, career perception of CS graduates was the second strongest predictor of students choosing to pursue a CS degree accounting for 27.5% of the varianc e in explainable factors. Results suggested that exposure to computer science courses in high school accounted for 22.4% of the variance in explainable factors. Lastly, Self - perception accounted for 17.1% of the variance in explainable factors affecting th e decision to pursue a CS undergraduate degree. Social encouragement included positive feedback from parents, teachers, and peers when pursuing an academic goal. Self - and perception of her pro ficiency in mathematics and problem - solving. Students exhibited this - (Google Inc. & Gallup Inc., 2014, p. 5) 21 Computer Science course in high schoo l, the existence of computing in the high school curricula, or access to computing - related extracurricular activities. Female students who had completed an Advanced Placement Computer Science course were found to be 38% more likely to pursue a Computer Sci ence degree. Lastly, career perception included having knowledge of the wide application of computer science beyond stereotypical views of CS as a solitary programming endeavor. Also included in career perception were the possible personal and professiona l benefits of a computing career. This was seen as not only combating media stereotypes of what a career in computing looks like, but also knowing that the work can have an effect on personal oogle & Gallup (2014) study also found that other factors, such as having a family member in the CS field, geography, early exposure to technology, and natural aptitude, had little or no influence on the likelihood for a student to enter into CS. With social encouragement, self - perception, academic exposure, and career perception being easier for te achers and parents to influence, these results were interpreted as a positive finding. The report suggested that these factors are malleable and recommended a number of steps for parents and educators. Considering that students in the United States do not formally choose a specialization until they reach university presents an opportunity to reach them at the secondary level or even earlier. The more a student has the opportunity to experience success in mathematics and computer science in high school, the likelier they are to continue in computer science as an undergraduate (Google Inc. & Gallup Inc., 2014, p. 2014) . While CS does not count towards high school graduation requirements in most states, twenty states have begun requiring schools to allow CS courses to count for some por requirements (Zinth, 2016) 22 and possibly a career path, coincides with these first exposures in high school (A. Lee, 2015; . This rise in support for CS as a college - preparatory course, coupled with research showing secondary level as the critical time for students to prepare for a CS pathway (Google Inc. & Gallup Inc., 2014, p. 2014; A. Lee, 2015; McInerney et al., 2006) , makes it increasingly important to examine whi ch of the current efforts show the greatest promise in increasing gender diversity in CS. High school is often one of the first settings in which students may take a course devoted entirely to the study of computer science, but many primary and middle scho ols are introducing programs such as the Hour of Code and Computer Science Unplugged to reach younger students (CS Education Research Group, 2014) . For example, Project Lead the Way has successfully offered professional development and curriculum for K - 8 teachers, thus preparing students for igh school (Brown, 2015) . These efforts prior to high school are part of an overall strategy in addressing inequity in CS, and provide an area for further study, yet this study is focused on the high school level due to the immediate availability of consistent CS curricula across a wide set of participating high schools. 2.4 Structural Barriers for Equity In addition to research d iscussed above on engaging female students in computer science, significant work has been done to understand how structural issues , such as disparate funding across educational systems , varied access to technology in the home, and prevalent gendered belief s, result in lack of access to STEM and CS offerings for historically marginalized groups (Goode, 2008; Goode, Margolis , & Chapman, 2014; Margolis, 2008, 2013) . Margolis (2008) examined three schools with varied resources and demographics in California to discover the root causes for differing gender, racial, and ethnic inequality in CS enrollment. The three 23 schools varied in the number of computing courses offered, the depth of the computing curriculum, and the computing resources available to students during non - class times. With funding disparitie s across the different schools, the authors found that computing courses were vulnerable to cuts as they were not a graduation requirement, the quality of computing equipment varied, and students had access to resources for longer hours in the more affluen t and teaching approaches that privileged students with prior experience with technology. The er students had a predisposed talent for computer programming, was especially problematic in that it appeared to be influenced by gender, race, and ethnicity of the students. The researchers also found that teachers would often privilege the knowledge and classroom interaction of students that already had prior experiences with technology, allowing their needs and questions to steer class discussion, thereby not giving all students equal opportunities and support to construct their own understanding of the material. Students also had unequal access to computers at home and stated their own racial and gendered students encountered sexist interactions with classmates in CS courses, teacher and student beliefs that male students had more inherent coding ability, and a lack of other female students to form a support network. The authors also found that the lack of computer science as a required course, such as mathemati cs or science, put the impetus upon teachers and students to recruit and maintain enrollment within computing courses. Some students and teachers were still successful personalities and resources. 24 In another study, Margolis and Fisher (2003) f ound that while both female and male students were motivated to major in CS as a result of enjoying programming, female students were also motivated by how computer science was applicable to other fields such as, science and math. This points to the need for CS curriculum to make these connections to other disciplines participation in computer science, there is limited research on what specific psychological fa ctors, such as belongingness, influence female students to pursue computing disciplines. Given the existing declines in overall enrollment of female students (U.S. Department of Education, 2017; Zweben, 2013; Zweben & Bizot, 2016) , additional research is needed to identify factors 2.4.1 Representations of Computing Disciplines In addition to the efforts discussed previously to broaden participation in CS, a growing body of work has also examined how CS curriculum can be more inclusive and showcase multiple ways in which computer scientists choose to work (Kafai & Burke, 2014; Kafai, Fields, & Searle, 2014; Searle, Fields, Lui, & Kafai, 2014; Turkle, 1997) . Turkle (1997) interviewed mu ltiple female programmers and found that traditional approaches to teach programming, reflected by a focus on linear problem solving and defining a program, were not reflective of the full range of programming styles. She argued that these traditional meth ods were a reflection of a male - dominated field. Turkle and Papert (1990, 1992) proposed the concept of programmers using a form of bricolage, which was not linear and required the programmer to pull from various resources and media, much like an artist; yet still resulting in the end with a program as valid as any other. The programmers using this method reported a somewhat chaotic process, not always driven by a clear plan, in which they jumped arou nd to work on code in various parts of a 25 program until they were satisfied with the overall result. This new type of work was between what one would call traditional programming, with a text editor and compiler, and design work previously done in an analog fashion. Part of this bricolage approach was that the contexts in which computing was applied were more than just on the screen, producing physical artifacts or products that were integrated with the computing technology. The authors saw this approach t o computing as a pathway to making the topic more relevant to students (Turkle & Papert, 1992) . Similar to Turkle and - traditional contexts for applying computing, Kafai and colleagues (Kafai & Burke, 2014, 2015; Kafai, Lee, et al., 2014; Kafai, Peppler, & Chapman, 2009) have done extensive work with physical computing involving textiles and circuitry, often with a focus on extending the appeal of CS to female and historically marginalized students. Students are exposed to new contexts for using computing skills by working with these physical projects, engaging in the social experience of programming, and expanding their conceptions of who can be a computer scientist. The rise of maker education (Blikstein, 2013; Dougherty, 2012; Halverson & Sheridan, 2014; Sheridan et al., 2014) has also introduced studen ts to computing ideas from a broader perspective. The results from these alternative approaches to learning and defining CS have pointed to greater participation and collaboration (Kafai, Fields, et al., 2014) , increased interest in computing (Searle et al., 2014) among female an d underrepresented minority students, and the development of computational thinking skills as a result of engaging in making (Wagh, Gravel, & Tucker - Raymond, 2017) Kafai et al. (2014) conducted a qualitative study with 15 high school students (7 female, 8 male) in a 10 - week physical computing module integrated into an existing computer science course. This module was comprised of lessons that used LilyPad Arduino circuits, a popular platform for programming that controls and interacts with sensors, lights, and other electronics. 26 These electronic components were integ rated into working with textiles to create an artifact that was presented at the conclusion of the module. The researchers reported that students found an increased sense of relevance for their computing skills, an increased ability to envision themselves as computer scientists, and an expanded idea of what computing tasks can include as valid work. The researchers also noted that girls and boys that were not typically attracted to computing were equally participating in the projects, while the use of a n on - competitive format for exhibiting final projects also increased engagement with computing concepts. Searle and Kafai (2015) also conducted a qualitative study looking at a three - week unit delivered as part of a female - only Native Studies course for American Indian girls aged 12 - 14 at a tribal charter school. The course focused on student - driven text ile projects that integrated circuitry with clothing (hoodies) that were shared at the end of the session. Students were encouraged to take the projects home with them to gain advice regarding sewing and crafting from knowledgeable friends and family. Th e researchers found that the focus on community and crafting of decorative clothing resonated with the cultural norms of the American Indians girls, increased their interest in computing, and gave them an expanded sense of their own capabilities. Wagh et a l. (2017) provided a case study of four 11th and 12th grade students (3 female, 1 male) engaged in making an interactive water fountain that reacted to musical tones with various lighting effects. The project was integrated with the use of Arduino circuit boards and LEDs, along with typical crafting supplies. Over the course of three weeks, the researchers repeatedly interviewed the participants in situ and documented the progression of the project via photographs and field notes. The team identified instances where the s tudents had naturally engaged in, and further developed, the computational thinking skills of problem decomposition, debugging, troubleshooting, and sense making. The researchers pointed to how the use of a 27 project personally important to students, the co llaborative nature of the project, and the instant feedback from code to LEDs allowed students to more effectively develop and use these computational thinking skills. These findings suggest that incorporating more of these non - traditional computing activi ties and approaches that go beyond typical programming courses into the CS curricula may help attract and retain students in computer science. This is supported by other research arguing that how we represent CS (Yadav et al., 2017) and student misconceptions about CS (Grover, Pea, & Cooper, 2014) has the potential to influence who participates in the field. I t is important that we represent the computer science di scipline in a way that broadens participation of traditionally underrepresented groups, including female students. 2.5 Context of the Study Computer science curriculum efforts in K - 12 are varied and difficult to match across schools, states, or regions. In a n effort to establish a more consistent curriculum across geographical and pedagogical divides, we chose to use the existing Advanced Placement Computer Science courses as a context for both curricular structure and sampling of students. 2.5.1 Advanced Placement Computer Science The AP Computer Science courses (AP Computer Science Principles and AP Computer Science A) are a set of courses offered in high schools throughout the United States which culminate in students taking a standardized test for each course. The College Board developed these courses with the input of both high school teachers and university faculty. The AP CS frameworks include a progression of topics and activities that teachers can rely upon. Considering the variation in how various states enact other computer science standards, the AP courses offer a n opportunity to sampl e students across geographical settings. Currently, there are 28 two AP CS courses Advanced Placement Computer Science A (APCS - A) and Advanced Placement Computer Science Pri nciples (APCS - P). The two courses reflect two differing approaches to the teaching of computer science the APCS - A course focuses primarily on Java programming while the APCS - P course takes a broad view of computing. This provides an opportunity for us to examine how a broad view of computer science influences student outcomes in CS. While specifics regarding our sample and location are covered in the methodology section, it is first important to understand the form and history of these two courses. 2.5.1.1 Advan ced Placement Computer Science A The College Board has offered APCS - A in some form since 1984. This course has historically been conceptualized and organized as a traditional programming course, with a focus on one specific language, programming lab exerc ises, and paper - based classroom tests. The style of work is largely individual in nature, although certainly instructors provide students the opportunity to work together at times. The course description provided by the College Board (2014) organizes the curriculum by programming constructs such as variables, methods, iteration, and classes. This course was intended to be the equivalent of an undergraduate introductory CS course (CS101) taken during the first semester (The College Board, 2014) , although APCS - A typically is done over the course of the entire academic year. The programming language, currently Java, is chosen by the College Board, with all official AP exams and materials reflecting only this language. The final APCS - A exam consists of questions regarding programming concepts and typical code - centric tasks, again in Java. 29 2.5.1.2 Advanced Placement Computer S cience Principles The College Board recently introduced the APCS - P course, partially in an effort to (The College Board, 2016a, p. 4) . This course was developed with the intent to broaden the appeal of computer science by focusing on s even big ideas of computing, including exposing students to computational thinking concepts and practices, and allowing students to examine how computing affects the world they live in (The College Board, 2016a) . The Seven Big Ideas of computing are creativity, abstraction, data and information, algorithms, programming, the internet, and globa l impact. The APCS - P curricular framework uses computational thinking throughout the course using six CT practices: connecting computing, creating computational artifacts, abstracting, analyzing problems and artifacts, communicating, and collaborating. A PCS - P was first officially offered to high school students in the 2016 - 2017 academic year. Similar to APCS - A, the APCS - P course was intended to be the equivalent of a first semester computing course at the undergraduate level. While the APCS - A course has also made some changes over the years to appeal to a larger audience of students, the APCS - P course was intended to be fundamentally different from APCS - A in its approach. Its content is not only new but contains more flexibility in its implementation an d format than the APCS - A course. The added that differ from th e traditional, solitary programming lab assignments. APCS - P provides non - programming exercises in the form of reports and group discussions about social issues tied to computing, such as conducting data collection of consumers and the larger implications of 30 doing so. Possibly the largest change, though, is that students are expected to create two projects The explore performance task involves researching a computing innov ation of the (i.e. virtual reality) innovation and its impact, and replying in written form to a series of prompts regarding their chosen topic. The artifact must be documented and s ubmitted as part of the AP exam. The create performance task involves the student creating a program to solve a problem of their own choosing. Students are heavily encouraged to work collaboratively with other students, and must use advanced logical and m athematical capabilities of their chosen programming language. The create performance task must be documented with video of the program running, written responses to prompts regarding their project, and submission of the final materials in an online AP di gital portfolio. These tasks are completed over the duration of the course and not intended as an end - of - course assessment. There is still a traditional end - of - course AP exam, but in place of using a prescribed language (e.g. Java), only pseudocode is prov ided so the student is not dependent on learning a specific language. Pseudocode is a term used to describe draft versions of a program written in natural language, not restricted to a formal programming language syntax, and used to describe the logical f low of a program without being concerned with exact structure and syntax. The remaining questions on the test are presented as traditional multiple choice word problems related to computer science principles and the relationship between computing and the world . In determining the final AP score for the course, which is graded on a 1 to 5 scale, the accounts for 24% of the score, and end - of - course exam accounts for 60% of the score. The 31 reduced emphasis on the final exam and the lack of a prescribed progra mming language are a reflection of the purpose of the course to broaden representations and contexts for the use of computer science. This expansion of the course con tent beyond only programming constructs, along with the expanded form of assessment beyond paper tests, is what distinguishes the APCS Principles course from the more traditional APCS - A course. The College Board intend s for the Principles cou rs e to aid wit h efforts to broaden participation in CS (The College Board, 2016a) . This is not the onl y goal of the Principles course, but a significant claim in College Board materials and related popular press (Anderson, 2018; The College Board, 2016a) . The course was chosen by the researchers as a setting not only for these claims, but more importantly for the curricular choices available in APCS - P that are tied to the research addressing gender inequity. The APCS - P framework requires students to apply computer science to authentic problems, outside of the context of a computer science of belonging in a computing classroom (Blaney & Stout, 2017; Searle et al., 2014; Searle & Kafai, 2015) . The use of computational thinking concepts adds to this cross - disciplinary reach of c omputer science , allowing for students to see the relevance of CS ideas to solving problems in other fields . While APCS - P is not a n all - inclusive representation of all the efforts to address gender inequity through curriculum, the inclusion of computationa l thinking, a broad representation of computer science, and addressing personally relevant problems are significant enough components to reasonably expect changes in student self - efficacy and belongingness. 32 2.6 Research Purpose Considering the two Advanced P lacement Computer Science courses and the intended purpose of APCS - P to broaden participation in computer science, this study address ed the following research questions: 1.) How do Advanced Placement Computer Science A and Advanced Placement Computer Science - efficacy, and persistence ? 1b.) How does student gender interact with the associations between course type and - efficacy, and persistence ? 2.) How do classroom gender proportions, and teacher gender in AP computer science courses - efficacy, and persistence ? 33 CHAPTER 3 : METHODS This chapter provides a summary of the methodology used for this study, including the sampling, participants, measures, procedures, and statistical analysis. The study was conducted with a correlational, predictive design (Creswell, 2008) , using cross - sectional survey methodology. The purpose of this design choice was to examine the correlation between - efficacy, and belongingness with their APCS course taken and their gender 3.1 Sampling Students enrolled in APCS - A and APCS - P were recruited through teachers who belonged to professional Computer Science or ganizations such as the Computer Science Teachers Association (CSTA), Michigan Association for Computer Users in Learning (MACUL), social media groups for AP CS instructors, and personal contacts. Both teachers and students were expected to participate in the study for a class section to be included in the analysis. Participants were spread across a diverse range of locations within the United States. The initial email to the teachers included a request to complete a preliminary survey about their teachin g experience and eligibility of classes to take part in the study. As this study included minors, teachers served as the primary conduit through which students were recruited, assent/consent materials collected, and survey links distributed. Using the re commendation from Kreft and de Leeuw (1998) regarding proper sample sizes for hierarchical linear modeling ana lysis, the researcher recruited at least 20 classes to participate in the study. Forty - eight teachers responded to the initial recruitment survey, of which 39 were eligible to take part in the study. Teachers were sent paper consent and assent forms for both the teachers (N=39) and the students (N=981) enrolled in their APCS classes. An additional 15 teachers and 34 their classes were dropped from the study for various reasons (withdrawal, non - response to emails, lack of consent, school policies), resulting in a total of 24 teachers (male = 12, female = 12). Concerns about including smaller class size s for hierarchical linear modeling as described by Maas and Hox (2004) led to the removal of classes with less than five students, with a final tot al of 17 teachers (8 male, 9 Female) and 20 class sections (10 APCS - A, 10 APCS - P). Of the 17 total teachers, 7 teachers taught APCS - A only, 7 teachers taught APCS - P only, and 3 teachers taught sections of both APCS courses concurrently. Specific number of student participants by gender for course sections are reflected in Table 1. Note that these are only study participants and do not reflect the total enrollment for each of the class sections. 35 Table 1 : Student Participant Summar y by Class Class ID Teacher ID Course Total Participants Male Participants Female Participants 1 1 APCS - A 5 3 2 2 2 APCS - A 10 6 4 3 2 APCS - P 27 19 8 4 3 APCS - A 11 7 4 5 4 APCS - P 6 3 3 6 5 APCS - A 8 8 0 7 6 APCS - A 7 2 5 8 7 APCS - A 19 14 5 9 7 APCS - P 14 8 6 10 8 APCS - P 14 10 4 11 9 APCS - P 9 7 2 12 10 APCS - A 5 5 0 13 11 APCS - A 22 17 5 14 11 APCS - P 48 37 11 15 12 APCS - P 5 5 0 16 13 APCS - P 12 4 8 17 14 APCS - A 10 8 2 18 15 APCS - P 17 10 7 19 16 APCS - A 6 6 0 20 17 APCS - P 8 6 2 Total 263 185 78 36 3.2 Participants Data was collected from a nationwide sample of 547 responses to the initial student survey. Participants were mostly from the Midwest region of the United States, with students from Arizona, Illinois, Indiana, Maryland, Michig an, New York, Ohio, Oklahoma, Pennsylvania, South Carolina, Texas, Vermont and Wisconsin taking part in the initial surveys . However, 242 responses were removed due to duplicate entries, attrition, lack of consent/assent, lack of a post - survey, enrollment in both CSA and CSP concurrently, and withdrawal. An additional 37 students were removed from analysis due to less than five participants within their respective individual class sections (Maas & Hox, 2004) . Given the focus of research on g ender differences, five students who did not provide their gender were also removed from the analysis. Final sample included 263 students with 185 male students and 78 female students. Students taking either of the two APCS course ranged from being in 9th to 12th grade, but the majority were in 11th or 12th grade, as are students in AP courses nationally (see Table 2 for detail ed demographics on participants). Racial demographics of the students (see Table 3) who self - identified were 189 White, 27 Asian, 2 2 multi - racial, 17 Black or African American, 2 Native American or Native Alaskan, and 6 as Other. There were 103 students in AP CSA course (27 females and 76 males) and 160 students in AP CSP course (51 females and 109 males). 37 Table 2 : Student Participant Summary by Gender and Grade Level Courses Grade 9 Male Grade 9 Female Grade 10 Male Grade 10 Female Grade 11 Male Grade 11 Female Grade 12 Male Grade 12 Female Total APCSA Only 9 2 6 1 34 13 27 11 103 APCSP Only 2 0 20 8 28 13 5 9 30 160 Total 11 2 26 9 62 26 86 41 263 Table 3 : Student Participant Summary by Race and Course Race All Total APCS A Total APCSP Male APCS A Female APCS A Male APCS P Female APCS P White 189 79 110 60 19 69 41 Asian 27 6 21 6 0 16 5 Multiple Race 22 6 16 2 4 12 4 Black or African American 17 8 9 5 3 8 1 Native American or Alaska Native 2 2 0 2 0 0 0 Other 6 2 4 1 1 4 0 Total 263 103 160 76 27 109 51 3.3 Measures 3.3.1 Student Surveys A survey was used to collect data on student (App endix A) demographic information, sense of belongingness in computer science, self - efficacy in computer science, and persistence 38 within computer science. Teachers also completed a survey (Appendix B) regarding demographics, computer science courses offere d at their school, courses they were currently teaching, and APCS class gender composition. Belongingness I version of the Math Sense of Belongingness Scale (Good et al., 2012) . The adapted survey included 28 items on an 7 - point Lik ert - type scale (Strongly Agree to Strongly Disagree), which were adapted for this study by changing item text portions from The scale was also adapted from Good et al. (2012) original 8 - point scale to a 7 - point scale to ensure all of the measures had consistent scale s , intended to avoid response error. Good et al. (2012) fou be ans. Good et al. (2012) for the composite belongingness score. For this subscales, Good et al. (2012) study, the subscale alpha ranged from 0.77 to 0.94 (see Table 4). 39 Table 4 : Cronbach's Alpha for Belongingness Scale Factor Good et al. (2012) Current Study Membership 0.95 0.92 Acceptance 0.91 0.94 Affect 0.91 0.94 Desire to Fade 0.78 0.88 Trust 0.81 0.77 Composite Belongingness 0.81 0.85 Self - effica cy The Self - Efficacy for Learning and Performance Scale, a component of the Motivated Strategies for Learning Questionnaire (MSLQ) (Pintrich, Smith, Garcia, & McKeachie, 1991) was used to assess st - efficacy for the course. The self - efficacy scale was comprised of eight items on a 7 - point Likert - type scale (see Appendix A), of which a mean of the responses was calculated to derive a self - efficacy composite score. Previously reported anal ysis of this 0.93) (Pintrich et al., 1991) . The scale has been used for assessing student self - efficacy in prior computer science education studie s (Lishinski et al., 2016) . For t his study internal validity was Persistence persistence in computer science, I adapted a measure from Yadav et al. (2014) original scale from Yadav et al (2014) was a four - point Likert - type scale, which I adapte d to a seven - point Likert - type scale for the purposes of this study to be consistent with other items in 40 student survey. The eleven items in the persistence section are Likert - type scale items that were interest in continuing t o study computer science topics, the value they place upon computer science in relation to their future career, and their interest in pursuing a career related to computer science. For this study internal validity of persistence scale was assessed with Cro Prior Grades unlikely and a source of further bias due to differing school records and GPA calculations for students. As a measure of prior academ ic achievement, we adapted a self - reporting scale from the National Education Longitudinal Study of 1988 (Ingels, 1990; NCES, 1988) . Research has established similar self - reported measures of academic achievement as acceptable and highly correlated with actual student grades (Frucgt & Cook, 1994; Kuncel, Credé, & Thomas, 2005) 3.3.2 Teacher Surveys An initial background survey was compl eted by teachers with items regarding their gender, race/ethnicity, courses offered at their school, courses they were expected to teach, and the gender composition of the CS course. A follow - up survey was completed after the semester had finished to verif y that they taught specific courses as initially expected, if any students had dropped the course, and any reasons students may have given for dropping the course. The first survey was composed of 17 items, while the second survey was composed of 8 items ( see Appendix B). 3.4 Procedure s Teachers were contacted via an email list for APCS - A and APCS - P instructors, personal contacts of the researcher, and social media posts for APCS teachers asking for their 41 participation and that of their students. Teachers comp leted an online recruitment survey to ensure that they were teaching at least one APCS course, expected class enrollment totals, and would start teaching the course in the fall semester. Teachers were compensated for taking part in the study. Students wer e compensated through a random drawing for Amazon gift cards if they completed all the forms and steps of the study. To avoid any concerns regarding coercion, students that chose to not take part in the study, yet returned a consent form stating so, were also included in the drawing pool for compensation. Compensation was sent directly to teachers, along with gift cards for students that were selected in the drawing. At the beginning of their fall semester, teachers that were teaching at least one APCS co urse in the fall semester were sent a collection of paper consent forms for their students to complete. Teachers were expected to collect and return these forms within the first two weeks of their course. As their courses began and consent forms were gat hered, teachers were then emailed a link to both the student pre - survey and initial teacher surveys. The teachers were encouraged to have students complete these surveys in a class setting to improve completion rate by early - September at the latest. Due to significant weather events (multiple hurricanes) that disrupted multiple schools, the deadline was extended to mid - September for all schools. The schools varied in their start dates for fall semester, from late July to early September, so priority was given to schools starting earlier when sending out the consent forms and links to surveys, with an effort to have each school receive these within the first week of class. The student pre - surveys gathered student background information, belongingness, sel f - efficacy, and persistence . - efficacy, and persistence , post - surveys were sent out in early December with a completion deadline of before 42 holiday break. Due to unforeseen school sch edule changes, weather, and unique school events, the deadline was extended into mid - January for these surveys as well. It is important to note that students were only halfway through their AP courses at the time when the post surveys were taken. They l ikely had enough experiences to see the influence of the curriculum on their belongingness and self - efficacy, as previous studies have shown (Cheryan et al., 2009; Lent et al., 1986; Lishinski et al., 2016; Master, Cheryan, & Meltzoff, 2016; Walton, Logel, Peach, Spencer, & Zanna, 2015) . Cheryan et al. (2009) and Master et al. (2016) which classroom or workplace décor was altered to be stereotypically male in nature. Lent et al. (1986) found difference in participants self - efficacy and vocational interest across a 10 - week course. These studies all point to the likelihood that any changes in self - efficacy and belongingness in this study can reasona bly be expected to take hold within one semester. During both the pre - survey and post - survey phases of the process, teachers were contacted via email to make them aware of how many students had completed their respective surveys, which consent forms had b een returned, and if their own teacher surveys were not yet - survey had passed in mid - January, preliminary analysis was conducted, and summary rosters sent to teachers for confirmation. Teachers replied when necessary with any rosters inaccuracies, such as a student choosing the incorrect course they were enrolled in, and any updates to course gender enrollment totals. 3.5 Analysis d environment variables. This presented class sections as a possible contextual variable, with it being likely that the sample of students would not be independent of each other and thus would 43 have student outcomes correlating within their class section. With students nested within each class section in the current study, three separate hierarchical linear modeling (HLM) analyses were used on the data to examine whether the predictive powers of APCS course and student gender on belongingness, self - efficac y, and persistence (each as an outcome in the model) varied between classes. The use of HLM accounts for the within - group variance for each class section, allowing for a more appropriate estimate of the effect of each variable than a simple multiple regre ssion on the entire collection of data would have provided. Intra - class correlation wa s calculated to determine if HLM wa s the appropriate statistical method to use for the study data. ICC could be interpreted as a percentage of the overall variance expl ained by the variance within the class sections, with recommendations from 0.02 to 0.10 being common thresholds to begin considering the use of HLM (Finch, 2014) over other methods. Multilevel analysis was conducted for two levels: variables related to the individual student (Level 1) and variables related to the class section (Level 2). The major analysis of the study was based on comparing student outcome s (belongingness, self - efficacy, and persistence ) between APCS - A and APCS - P, while adjusting for differences between clusters on Level - 1 (individual student) and Level - 2 (class section) covariate characteristics. Level - - t est levels of belongingness, self - efficacy, and persistence, along with student gender, and self - reported prior academic performance. Level - 2 covariates included teacher gender and class gender balance percentages. 44 CHAPTER 4 : RESULTS This chapter describes the ana lysis and results conducted on the resulting data from the student and teacher surveys described in Chapter 3. 4.1 Descriptive Statistics Summary statistics (Table 5 ) and correlations (Table 6 ) for self - efficacy, belongingness, persistence, and student prio r grades can be found below. Table 5 : Descriptive S tatistics for Self - Efficacy, Belongingness, Persistence, and Student Prior Grades Self Efficacy Pre Self Efficacy Post Belong. Pre Belong. Post Persist. Pre Persist. Post Student Grades N 263 263 263 263 263 263 263 Median 5.88 5.88 5.45 5.50 5.55 5.64 6.00 Mean 5.71 5.62 5.43 5.46 5.47 5.41 5.67 SD 1.08 1.15 0.86 0.88 1.16 1.25 0.62 Minimum 2.00 1.38 3.33 2.85 1.64 1.36 1.00 Max 7.00 7.00 7.00 7.00 7.00 7.00 6.00 Pearson correlations provide a preliminary indicator of how the pre and post scores for self - efficacy, belongingness, and self - efficacy may be related. As such, the reader should note that subsequent analysis using hierarchical linear modeling will provi de more accurate results of Results suggested that pre self - efficacy had positive correlations with post self - efficacy ( r =0.68, p<0.001) and student self - reported prior grades ( r =0.20, p<0.01). Post self - efficacy had a positive correlation with student self - reported prior grades ( r =0.14, p<0.05). Pre belongingness had a positive relationship with post belongingness (r=0.64, p<0.001). Results suggested that pre persistence had a positive correlation with post persistence (r=0.73, p<0.001). 45 Table 6 : Self - Efficacy, Belongingness, Persistence, and Student Prior Grades Correlations Self Efficacy Pre Self Efficacy Post Belong. Pre Belong. Post Persist. Pre Persist. Post Self Eff. Pre Self Eff. Post 0.68*** Belong. Pre 0.64*** 0.46*** Belong. Post 0.45*** 0.60*** 0.64*** Persist. Pre 0.40*** 0.30*** 0.47*** 0.32*** Persist. Post 0.30*** 0.40*** 0.32*** 0.49*** 0.73*** Student Grades 0.20** 0.14* 0.02 - 0.01 0.06 0 .07 Notes * p<0.05 ** p<0.01 *** p<0.001 As shown in Table 7 , APCS - P courses had a higher proportion of female participants (female = 31.9%, male = 68.1%) than APCS - A courses (female = 26.2%, male = 72.8%). A chi squared test of independence was conducted to examine whether there was a relationship between the course taken and student gender. The relationship between these variables was not significant ( 2 (1) = 0.71, p=0.40). It is important to keep in mind that these percentages are for total participants in the study, which is distinct from the class gender percentages reported by teachers which included both study participants and non - participants. 46 Table 7 : Frequency of Student Participants by Course and Gender Female Students Male Students APCS - A 27 (26.2%) 76 (72.8%) APCS - P 51 (31.9%) 109 (68.1%) Total 78 (29.7%) 185 (70.3%) As shown in Table 8 , participants from APCS - P courses were more likely to have a female teacher (female = 52.5%, male = 47.5%) than APCS - A courses (female = 32.0%, male = 70.0%). A chi squared test of independence was conducted to examine whether there was a relationship between the course taken a nd teacher gender. The relationship between these variables was significant ( 2 (1) = 9.8, p=0.002), with students in APCS - P being more likely to have a female teacher than students in APCS - A. Table 8 : Frequency of Teacher Gender b y Course Students with Female Teacher Students with Male Teachers APCS - A 33 (32.0%) 70 (68.0%) APCS - P 84 (52.5%) 76 (47.5%) Total 117 44.5%) 146 (55.5%) As shown in Table 9 , APCS - P classes had a higher proportion of female teachers (female = 60.0%, male = 40.0%) than APCS - A courses (female = 40.0%, male = 60.0%). A chi squared test of independence was conducted to examine whether there was a relationship between the course taken and teacher gender. The relationship between these variables was not fo und to be significant ( 2 ( 1 ) = 0.2 , p= 0.655 ). 47 Table 9 : Frequency of Teacher Gender for Each Class Section by Course Type Class Sections with Female Teacher Class Sections with Male Teachers APCS - A 4 (40.0%) 6 (60.0%) APCS - P 6 (60.0%) 4 (40.0%) Total 10 (50.0%) 10 (50.0%) 4.2 Outcome Variable Change To analyze pre - post changes in each outcome variable, I calculated a pre - post change for self - efficacy, belongingness, and persistence. This was calculated by subtracting pre scores from post scores for each of the three outcome variables. Table 10 (see below) offers a summary of the self - efficacy post scores by student gender, course taken, and teacher gender to give an overview of how the different factors of interest compare. Of note is the relatively small mean change in self - efficacy for all participants (M= - 0.10, SD=0.89). Female (M= - 0.19, SD=0.96) and male (M= - 0.05, SD=0.86) students both showed a minor decrease in self - efficacy. APCS - A students show a minor increase in sel f - efficacy (M=0.02, SD=1.01), while APCS - P students show a minor decrease in self - efficacy (M= - 0.17, SD=0.79). Students with both female teachers (M= - 0.08, SD=0.86) and male teachers (M= - 0.11, SD=0.91) showed a decrease in self - efficacy. 48 Table 10 : Descriptive Statistics for Self - Efficacy by S tud ent Gender, Course, and Teacher Gender N Self - Efficacy Pre Self - Efficacy Post Self - Efficacy Change Mean SD Mean SD Mean SD Student Gender Female 78 5.31 1.12 5.12 1.32 - 0.1 9 0.96 Male 185 5.88 1.02 5.83 1.00 - 0.05 0.86 Course APCS - A 103 5.43 1.18 5.46 1.25 0.02 1.01 APCS - P 160 5.89 0.98 5.72 1.07 - 0.17 0.79 Teacher Gender Female 117 5.63 0.97 5.55 1.10 - 0.08 0.86 Male 146 5.77 1.16 5. 67 1.19 - 0.11 0.91 Total 263 5.71 1.08 5.62 1.15 - 0.10 0.89 Table 1 1 (see below) offers a summary of the Belongingness post scores by student gender, course taken, and teacher gender to give an overview of how the different factors of interest compare. Of note is the relatively small mean change in belongingness for all participants (M=0.02, SD=0.74). Both female (M=0.04, SD=0.63) and male (M=0.02, SD=0.78) students experienced a minor increase in belongingness. APCS - A students showed a minor gain in b elongingness (M=0.12, SD=0.76) while APCS - P students showed a minor decrease (M= - 0.04, SD=0.72). Students in classes with female teachers experienced a very minor decrease in belongingness (M= - 0.01, SD=0.69), while those in classes with male teachers show ed a minor increase (M=0.02, SD=0.748. 49 Table 11 : Summary Statistics for Belongingness by Student Gender, Course, and Teacher Gender N Belongingness Pre Belongingness Post Belongingness Change Mean SD Mean SD Mean SD Student Gender Female 78 5.16 0.84 5.20 0.9 0.04 0.63 Male 185 5.54 0.85 5.56 0.84 0.02 0.78 Course APCS - A 103 5.27 0.91 5.39 0.94 0.12 0.76 APCS - P 160 5.54 0.81 5.50 0.83 - 0.04 0.72 Teacher Gender Female 117 5.34 0. 86 5.34 0.89 - 0.01 0.69 Male 146 5.50 0.86 5.55 0.85 0.05 0.78 Total 263 5.43 0.86 5.46 0.88 0.02 0.74 Table 1 2 (see below) offers a summary of the Persistence post scores by student gender, course taken, and teacher gender to give an overview of h ow the different factors of interest compare. Of note is the relatively small mean change (M = - 0.06, SD = 0.90) in persistence for all participants. Female students showed a minor positive change in persistence (M = 0.02, SD = 0.78) while male students showed a minor decrease (M = - 0.10, SD = 0.94). APCS - A students also showed a minor increase in persistence (M=0.02, SD=0.72) while APCS - P students showed a minor decrease (M= - 0.10, SD=0.99). There was a minor decrease for students in classes with both f emale teachers (M= - 0.08, SD=0.73) and male teachers (M= - 0.05, SD=1.01). 50 Table 12 : Summary Statistics for Persistence by Student Gender, Course, and Teacher Gender N Persistence Pre Persistence Post Persistence Change Mean SD Mean SD Mean SD Student Gender Female 78 5.14 1.26 5.17 1.34 0.02 0.78 Male 185 5.61 1.09 5.51 1.20 - 0.10 0.94 Course APCS - A 103 5.54 1.19 5.55 1.19 0.02 0.72 APCS - P 160 5.43 1.14 5.32 1.29 - 0.11 0.99 Teacher Gender Female 117 5.22 1.20 5.15 1.32 - 0.08 0.73 Male 146 5.67 1.09 5.62 1.16 - 0.05 1.01 Total 263 5.47 1.16 5.41 1.25 - 0.06 0.90 4.3 Normality of Outcome Variables An area of concern was the normality of the outcome variables, an assumption of the use of hierarchical linear modeling. As seen in Table 1 3 , the calculated standardized skewness statistics for Self Efficacy Post ( - 6.215), Belongingness Post ( - 2.923), and Persistence Post ( - 5.506) showed notable negative skew. This negative sk ew reflect ed the scores having been consistently recorded toward the upper end of the 7 - point Likert - type scales for each of the measures for self - efficacy, belongingness, and persistence. A square transformation was performed to bring the outcome variables within more acceptable ranges for analysis. 51 Table 13 : Outcome Variable Skew and Kurtosis Self - Efficacy Post Belong. Post Persist. Post Self - Efficacy Post Squared Belong. Post Squared Persist. Post Squared N 263 263 263 263 263 263 Sk ewness - 0.934 - 0.439 - 0.827 - 0.349 - 0.028 - 0.260 SE of Skewness 0.150 0.150 0.150 0.150 0.150 0.150 Kurtosis 0.894 - 0.045 0.406 - 0.661 - 0.549 - 0.867 SE of Kurtosis 0.299 0.299 0.299 0.299 0.299 0.299 Skew Std. - 6.215 - 2.923 - 5.506 - 2.326 - 0.186 - 1.728 Kurtosis Std. 2.987 - 0.151 1.357 - 2.208 - 1.834 - 2.897 4.4 Analysis and Results The analyses for the effects of student gender and course type (APCS - A vs. APCS - P) on self - efficacy, belongingness, and persistence, using hierarchical linear modeling, are cover ed in this section. A separate analysis was conducted for each of the outcome variables: self - efficacy, belongingness, and persistence. 4.4.1 Self - Efficacy Analysis and Results A null model with self - efficacy post scores, with intercepts allowed to vary by cla ss section, was examined to determine if hierarchical linear modeling (HLM) was necessary for analysis. The intraclass correlation (ICC) was found to be 0.079, which suggested that 7.9% of the variance in self - efficacy was between class sections. An ICC of this amount in hierarchical data is considered sufficient to justify using hierarchical linear modeling methods (Niehaus, Campbell, & Inkelas, 2014) . Level One variables were self - effica cy pre scores, student gender, and student prior grades. Level Two variables were teacher gender, APCS course taken, and percentage of male students in the class. 52 Self - efficacy pre scores, student prior grades, and class percentage of male students wer e centered and standardized to reduce any possible issues with multicollinearity (Finch, 2014) , using grand mean centering due to the course type (Level Tw o) variable being of primary interest (Field, Miles, & Field, 2012) . All models used maximum likelihood method for gen erating parameter estimates. In step two of our analysis, the model was further developed by individually adding the Level One predictors (self - efficacy pre scores, student gender, student prior grades) as fixed effects. The addition of both self - efficacy improvement in model fit, yet was retained in the model as theoretically important as a covariate. In step three, each of the Level Two variables (teacher gender, APCS course taken, and percentage of male student in the class) were added as static slope, with none of the variables signifi cantly improving the fit of the model. Similar to how the Level One variables were retained, the Level Two variables were retained as they were theoretically important for the research questions and did not significantly worsen the model fit. In step fou r, each of the Level One variables (self - efficacy pre scores, student gender, student prior grades) were tested individually as a random slope, yet none of the variables produced a statistically significant improvement to the model fit. Testing these vari ables as a random slope allowed me to test whether the model is improved by allowing the variable to have a different slope value for each of the class sections. As a result of the lack of model improvement , all the Level One variables remained as fixed s lope variables in the model. In step five, each Level Two variable was tested as a random slope, just as the Level One variables were, yet none of the variables produced a statistically significant improvement to the model fit. As a result, all the Level Two variables 53 remained as fixed slope variables in the model. In the final step, an interaction effect was tested between student gender and teacher gender, with the model at this point being unable to converge. The inability of the model to converge su ggests that this model was uninterpretable and unusable model. The resulting best fit model from this process used only fixed slopes for the predictor variables, but did allow the intercept to be random across classes. The model was as follows, Final S elf - Efficacy HLM Model: Level 1: Y ij 0j 1 j(Se l 2j 3j 0j 00 01 02 03 (PercentMale) + u 0j Mixed Model: Y ij 00 01 02 03 (PercentMale) + 10 (SelfEffPre) + 20 (StudentGender) +y 30 (StudentGrades) + u 0j +u 2j Effects found include a positive relationship for self - efficacy post scores with self - efficacy pre - 0 .11, p = 0.02, ES = 0.02). This suggests a medium to large effect size for the self - efficacy pre scores on the self - efficacy post score. A significant effect from student gender on self - efficacy post - scores can be interpreted as female students experienci ng a decline in self - efficacy over the course of APCS course taken, and class percentage of male students. The resulting statistics are shown in Table 1 4 . 54 Table 14 : Self - Efficacy Hierarchical Linear Model Analysis Self - Efficacy Post B std. Beta SE p Fixed Parts (Intercept) 34.30 <.001 Self - Efficacy Pre 7.64 0.65 0.05 <.001 Female Student - 2.89 - 0.11 0.05 .020 St udent Grades 0.41 0.03 0.05 .490 Female Teacher - 0.18 - 0.01 0.06 .900 APCS - Principles - 0.44 - 0.02 0.06 .775 Class % Male 0.19 0.02 0.06 .792 Random Parts 2 70.232 00, classID 2.808 N classID 20 ICC classID 0.038 Observations 263 R 2 0 2 .510 / .509 4.4.2 Belongingness Analysis and Results A null model with belongingness post scores with intercepts allowed to vary by class section, was examine d to determine if hierarchical linear modeling (HLM) was necessary for analysis. The intraclass correlation (ICC) was found to be 0.026, suggesting that 2.6% of the variance in 55 belongingness was between class sections. An ICC of this amount in hierarchica l data is considered sufficient to justify using hierarchical linear modeling methods (Niehaus et al., 2014) Level One variables were belongingness pre scores, student gender, and student prior grades. Level Two var iables were teacher gender, APCS course taken, and percentage of male students in the class. Belongingness pre scores, student prior grades, and class percentage of male students were centered and standardized to reduce any possible issues with multicoll inearity (Finch, 2014) , using grand mean centering due to the course type (Level Two) variable being of primary interest (Field et al., 2012) . All models used maximum likelihood method for generating parameter estimates. In step two of our analysis, the model was further developed by addin g the Level One predictors (belongingness pre scores, student gender, student prior grades) as fixed (3) = 153.07, p < 0.001) from the previous model. Studen reported did not show a significant change in the model fit, but were retained in the model as theoretically important. In step three, each of the Level Two variables (teacher gender, APCS course taken, and percentage o f male student in the class) were added as static slope, with none of the variables significantly improving the fit of the model. Similarly to how the Level One variables were retained, the Level Two variables were retained as they were theoretically imp ortant for the research questions and did not significantly worsen the model fit. In step four, each Level One variable was tested as a random slope, with the finding that allowing student gender to be random produced a statistically significant improveme 6.1923, p < 0.032). However, the correlation of the random slope of student gender ( - 1.00) denoted a possible problem with an overfit model , providing a model that was too closely tied to 56 the data collected to be generalized to other data sets . Testing for singularity confirmed the overfit fit model, so student gender was not made random and all Level One predictors remained static. In step five, each Level Two variable was tested as a random slope, yet none of the variables produced a statistically significant improvement to the model fit. As a result, all the Level Two variables remained as fixed slope variables in the model. In the final step, an interaction effect was tested between student gender and teacher gender, bu t failed to find a significant difference. The resulting best fit model from this process used fixed slopes for all of the predictor variables except student gender, which along with the intercept was allowed to be random across classes. The model was as follows: Final Belongingness HLM Model: Level 1: Y ij 0j 1 2j 3j 0j 00 01 02 03 (PercentMale) + u 0j Mixed Model: Y ij 00 01 02 03 (PercentMale) + 10 20 (StudentGender) +y 30 (StudentGrades) + u 0j + u 2j Effects found include a positive relationship for belongingness pre scores with belongingness bel ongingness pre scores on the belongingness post score. The model failed to show any class percentage of male students. The resulting statistics are shown in T able 1 5. 57 Table 15 : Belongingness Hierarchical Linear Model Analysis Belongingness Post B std. Beta SE p Fixed Parts (Intercept) 31.44 <.001 Belongingness Pre 6.08 0.66 0.05 <.001 Female Student - 0.60 - 0.03 0. 05 .539 Student Grades - 0.17 - 0.02 0.05 .707 Female Teacher - 0.77 - 0.04 0.06 .520 APCS - Principles - 0.84 - 0.04 0.07 .507 Class % Male 0.47 0.05 0.06 .446 Random Parts 2 45.550 00, classID 2.331 N classID 20 ICC classID 0.049 Observ ations 263 R 2 0 2 .483 / .483 4.4.3 Persistence Analysis and Results As the first step of analysis, a null model with persistence post scores with intercepts allowed to vary by class section, was examined to determine if hierarchical linear modeling (HL M) was necessary for analysis. The intra class correlation (ICC) was found to be 0.149, suggesting that 14.9% of the variance in persistence was between class sections. An ICC of this amount in hierarchical data is considered sufficient to justify using h ierarchical linear modeling methods (Niehaus et al., 2014) . 58 Level One variables were persistence pre scores, student gender, and student prior grades. Level Two variables were teacher gender, APCS course taken, and percentage of male students in the class. Persistence pre scores, student prior grades, and class percentage of male students were centered and standardized to reduce any possible issues with multicollinearity (Finch, 2014) , using grand mean centering due to the course type (Level Two) variable being of primary interest (Finch, 2014) . All models used maximum likelihood method for generating parameter estimates. In step two of our analysis, the model was further developed by adding the Level One predictors (persistence pre scores, student gender, student prior grades) as fixed slopes, with the show a significant change in the model fit, but were retained in the model as theoretically important. In step three, each of the Level Two variables (teacher gender, APCS course taken, and percentage of male student in the class) were added as static slope, with none of the variab les significantly improving the fit of the model. Similarly to how the Level One variables were retained, the Level Two variables were retained as they are theoretically important for the research questions and did not significantly worsen the model fit. In step four, each Level One variable was tested as a random slope, yet none of the variables produced a statistically significant improvement to the model fit. As a result, all the Level One variables remained as fixed slope variables in the model. In step five, each Level Two each Level Two variable was tested as a random slope, yet none of the variables produced a statistically significant improvement to the model fit. As a result, all the Level Two variables remained as fixed slope variables in the model. In the final step, an interaction effect was tested between student gender 59 and teacher gender but failed to find a significant difference. The resulting best fit model from this process used fixed slopes for all the predictor variables, while the intercept was allowed to be random across classes. The model was as follows: Final Persistence HLM Model: Level 1: Y ij 0j 1 2j 3j 0j 00 01 02 03 (PercentMale) + u 0j Mixed Model: Y ij 00 01 02 03 (PercentMale) + 10 (PersistPre) + 20 (StudentGender) +y 30 (StudentGrades) + u 0j +u 2j Effects found include a positive relationship for persistence pre scores with persistence post scores on the persistence post score. The model failed to show any significance for student students. The resulting statistics are shown in Table 1 6 . 60 Table 16 : Persistence Hierarchical Linear Model Analysis Persistence Post B std. Beta SE p Fixed Parts (Intercept) 28.46 <.001 Persistence Pre 7.70 0.72 0.04 <.001 Female Student 0.76 0.03 0.04 .528 Student Grades 0.19 0.01 0.04 .824 Female Teacher - 0.92 - 0.04 0.04 .398 APCS - Principles - 0.92 - 0.04 0.04 .417 Class % Male 4.53 0.05 0.05 .264 Random Parts 2 70.039 00, classID 0.000 N classID 20 ICC classID 0.000 Observations 263 R 2 0 2 . 545 / .545 61 CHAPTER 5 : DISCUSSION This study examined how two different Advanced Placement Computer Science courses, APCS - A and APCS - Principles, affect - efficacy, belongingness, and persistence within computer science, and how those relation ships relate to student gender. The three outcome variables of self - efficacy, belongingness, and persistence were chosen as possible indicators of the likelihood of students to pursue computer science. The relationship between student gender and the two courses was examined due to the underrepresentation of women in computer science. Teachers and students from a nationwide sample of APCS courses completed multiple surveys to provide a rich data set. Hierarchical linear modeling analysis was conducted to account for the nested nature of the students within their respective class sections. Based upon also included in the analysis. Hierarchical linear analysi s was conducted with separate models for each of the outcome variables (self - efficacy, belongingness, and persistence). - efficacy scores and student gender significantly predicted post self - efficacy scores. The APC gender, and class percentage of male students did not have a significant relationship with post self - efficacy scores. Student gender had a small, negative effect on post self - efficacy scores, which suggested that female students had a larger decrease in self - efficacy than male students over the course of the study. These results match with existing research, which has found that male students take longer than female students to match their self - efficacy beli efs with actual task performance (Beyer, 2014; B eyer, Rynes, Perrault, Hay, & Haller, 2003) self - efficacy is influenced by their performance in a CS course (Lishinski et al., 2016) . Resea rch has also suggested that even in typically female - majority courses such as biology, male students 62 have a higher sense of confidence in their skills than female students with similar grades (Cooper, Krieg, & Brownell, 2018) . e belongingness scores were significantly related gender, and class percentage of male students did not have a significant relationship with belongingness post scores. While the effect of course chosen and student gender are the main focus of this study, I found the lack of influence of two covariates (teacher gender and class gender percentages) on belongingness to be surprising. The lack of a change in belong ingness, particularly in the APCS - P course , was surprising as this course was supposed to engage traditionally underrepresented students in computing ideas, including women , to broaden the participation of students in computer science. No difference was f ound by gender for abilities in computer science (Beyer, 2014; Cundiff et al., 2013; C. M. Lewis, Anderson, & Yasuhara, 2016; Master et al., 2016) and female students in this study were in minority in their classes. Results from prior research related to stereotype threat (Cheryan et al., 2009; Murphy et al., 2007; Steele, 1997) would suggest that teacher gender and class percentage of student gender would have a significant e ffect on student belongingness, particularly for female students. - scores were prior grades, teacher gender, and class percentage of male students did not have a significant relationship with persistence post scores. Considering how exposure to computer science courses is shown in prior research to be a significant factor in whether students choose to major 63 in computing fields (Google Inc. & Gall up Inc., 2014) , and how APCS - Principles was intended to improve upon female students interest in continuing in CS , these results are surprising. One possible explanation for the lack of APCS course predicting self - efficacy, belongingness, or persistence might be due to the enactment of the two APCS courses and inherent variability in how teachers implemented the curriculum. Only knowing the course designation does not provide information on whether the CS principles included a broad representation compute r science and how computing is applicable within other disciplines. Prior research has suggested that female students are attracted to a broader view of computer science and how what they learn in CS can b e used to solve problems in other disciplines (Margolis, Fisher, & Miller, 2000) . Future work should examine how different instantiations of the APCS - P - efficacy, belonging, and persistence . Future work could also gather data describing the specific pedagogical strategies teachers used within the APCS - P and APCS - A classes and examine whether these had an effect on belongin gness or self - efficacy in computer science. Within the range of possible instantiations and pedagogical practices used in APCS - A or APCS - P, there are likely examples that do influence form, the study did not cap ture these distinctions within the range of courses sampled . This examination of practices in the classroom should be done from both from a student and teacher perspective, as while teachers may fully intend to use different approaches, students may not r eceive or experience these efforts in the way it was intended. and pedagogical preferences. This study did not collect data related to the actual practices taking part in the classroom, or attempt to capture teacher beliefs that relate to their pedagogical choices. If these beliefs and choices were independent of which courses they taught, and thus 64 affected both courses equally, it would explain why the results by course were so similar and no difference by course was found. Another possible hypothesis for the lack of findings could be due to the diffusion of practices between APCS - A and APCP - P. The m ajority of participating APCS teachers had taught or were current ly teaching other computer science courses, with a significant portion having taught both APCS - A and APCS - P. It is very likely that as they attended professional development workshops, reviewed professional teaching materials, and engaged with the CS teac hing community, that teachers of both APCS - P and APCS - A integrated some of these recommended practices for making their courses more equitable. As a result, APCS - A students may have benefited from these equity efforts, and thus explain the lack of differen ce in results across courses. The inverse could also be true, in that newly learned practices related to equity were not fully enacted into either of the courses, with not enough difference to be found between APCS - A and APCS - P as a result. Again, we do n ot have enough information about what occurred within the classroom, particularly the pedagogical choices being made and the classroom social interactions, to be able to say how present or not these practices were. Considering the lack of significant chang - efficacy, belongingness, and persistence found in this study, a possible explanation is that students entering into the APCS courses have already formed their beliefs around their relationship to computer science. Participants were lar gely juniors and seniors in high school, which suggests we need to reach them before high school or even earlier to allow them to see relevance of CS to other fields. This aligns with suggestions by Grover et al. (2014) and Shapiro et al. (2015) that middle school aged and younger students may be a more appropriate age group to target for addressing gender inequity in computer science. Students have already received and are beginning to process 65 messages regarding the gendered notions of computer science by middle school (Shapiro et al., 2015; Yadav et al., 2017) . Research shows tha t primary and middle school students have successfully developed and used computational thinking concepts (Grover et al., 2014) . While researchers focusing on younger age levels would lose the common curricular guidance of the AP programs, it is possib le that other curricular efforts through existing organizations such as Project Lead the Way, Code.org, or Google could provide a similar consistent, national framework. In considering this shift to younger students, it bears considering that this focus on earlier grades does not release us as researchers and educators from our responsibility to continue addressing gender inequality in the upper grade levels. Finally, the lack of change in self - efficacy, belongingness, and persistence may be that the measur es adapted for this study were too general and not nuanced enough to measure constructs within computer science. The field of computer science is referred to broadly in these mputer (2017) similarly hypothesized that these general measures are no t detailed enough for the domain, and that instruments that relate to sub - concepts of computer science could be developed that could be better understood by students. Rather than make the instruments could be adjusted with more approachable language for CS novices. Additionally, this is an area where science is prior to and after working within the course. 5.1 Implications The results found that the type of AP Computer Science course did not significantly influence - efficacy, belongingness, and persistence. As such, the implication is that the 66 current implementation of APCS Principles cou rse may not be associated with an increase in 5.1.1 Implications for Practice For those teaching computer science at the high school level, or involved in developing the AP CS curriculum, these may not appear to be helpful results at first glance. Consider that female students accounted for only 23.6% of APCS - A enrollees during the 2016 - 2017 academic year, while in APCS - P female students accounted for 30.1% of APCS - P enrolle es. While this shows an improvement in the recruitment of women into CS courses, this proportion is still very low and considerable work still needs to be done to achieve comparable gender proportions in computer science. This suggests a need to look beyo nd high school curriculum alone to address gender inequity, possibly examining structural barriers outside of the classroom that determine who is more likely to walk into the classroom door. School administrators, counselors, and policymakers can help ad dress structural barriers by looking for instances where students are not currently being directed toward computer science and asking why that is. While APCS - P was intended to address some of these issues, it is possible that we are still mostly bringing k ids in the door that would have likely ended up in computer science anyway. Efforts like CSforAll, which are focusing on exposing all students to some computer science, without needing to be enrolled in a formal CS course , may help with the lack of CS exp erience. The integration of computational thinking approaches into non - CS courses may also aid in exposing students to the idea that they are already doing CS types of activities in their existing work (Benakli, Kostadinov, Satyanarayana, & Singh, 2017; Hambrusch, Hoffmann, Korb, Haugan, & Hosking, 2009; I. Lee, Martin, & Apone, 2014; Voogt, Fisser, Good, Mishra, & Yadav, 2015; Yada v, Hong, & Stephenson, 2016) . This not only reaches 67 students with CS while avoiding the need to be in a CS course, but can also provide authentic self - efficacy t o pursue computer science and feel like they belong. In addition, introducing CS ideas by integrating in core subject areas could expose teachers and administrators to the range of possible applications of computer science, and adjust their conception of who would benefit from taking a course in computer science. An additional consideration for future practice for administrators would be to examine whether they are assigning teachers of differing experiences to the different types of computing courses. It is possible that if there is a disparity in teaching experience, that disparity is detracting from possible benefits to be had from the existing curriculum. APCS - P by design is more approachable for novice students in programming (The College Board, 2016a) and that may result in teachers with less programming experience to be more commonly as signed to teach APCS - P courses. Future practice and research can examine the exact practices put into use by the teachers, whether the practices correlate with years of experience or other measures of content knowledge. There may also be biases present in which schools are able to offer more advanced computing courses, such as APCS - P and APCS - A, due to disparate funding and staffing (Margolis, 2008) The results from this study - efficacy was negatively affected to a larger degree than their male counterparts, similar to the findings of Lishinski et al. (2016) . A s suggested by Lishinski et al., educators should be carefully scaffolding students experiences to ensure that they are sense of self - efficacy, but not so challenging that they are discouraging stu dents . 68 5.1.2 Implications for Future Research A consideration for future research is developing measures that are applicable to computer science. We need to develop measures that are sensitive enough to measure self - efficacy, belongingness, and persistence wit hin computer science. In addition, we also need to better understand how students conceive of computer science. Future research needs to examine - efficacy, belongingness, persistence . We ha ve little sense of how to interpret changes in student self - efficacy, belongingness, or persistence without knowing more about their reasons for their answers. Future research could also examine how other variables such as gendered student perceptions of CS, as well as teachers beliefs about who can do CS, could also influence - efficacy, belongingness, persistence . Shumow and Schmidt (2013) found that teacher beliefs in a science classroom can lead to higher levels of classroom interaction between male students and instructors, resulting in inequita ble learning experiences for female students. Reigel - Crumb and Humphries (2012) gendered beliefs regarding the mathematic ability was related to more negative evaluations of female students. Espinoza et al. (2014) found that mathematics teachers were more likely to that interventions could help address these beliefs and the influence on classroom practice. Future research could use examine similar phenomena in computer science settings, with classroom obse rvation to see how teachers interact with students in the classroom and how their beliefs influence their pedagogical choices. In addition to the effect of teacher beliefs, student beliefs in fixed ability traits, or student perceptions of the prevalence of ability beliefs among peers, could have an effect on 69 belongingness. When female students perceive that teachers or peers believe that ability to learn a particular discipline is innate and fixed, there is a negative effect on their sense of belongingn ess (Good et al., 2012) . Consider that a female student m ay enter a classroom and see a disproportionate number of male students, and then encounters statements about the need for innate talent in this field . T his could arguably lead to her developing a lack of belonging in the discipline. Smith et al. (2013) found that female students in male dominated STEM fields often believed that they will have to work harder than their male counterparts, interpreted this as a lack of ability, and resulted in a lower level of motivation. We did not e ability beliefs in this study, and thus may have missed a significant predictor of existing belongingness beliefs. Future research can examine these beliefs in computer science classrooms, along with perceptions students have of As more time passes from the introduction of the APCS - P courses, we can more directly study the relationship between retention of female and other underrepresented students within computer science by examining college acceptance, dropo ut, and graduation rates for students that took the APCS courses in high school. However, this requires longitudinal studies to examine the impact of APCS - P on persistence in CS. There also needs to be work on examining experiences of other underreprese nted groups, including African - American and Latino students in computer science. However, typical low enrollment of these groups in APCS courses requires national sampling of participants. In addition, future research could also use qualitative approaches to complement our understanding of why students choose to persist (or not) within computer science, and the effect that courses like APCS have on them. Using qualitative methods to more fully understand the process by which students make their decisions, and the cultural and gendered messages they receive in 70 relation to the course could also inform development of curriculum and pedagogical practices used to engage traditionally underrepresented groups in CS. 5.2 Limitations The limitations for this study are l argely the result of the chosen study methodology and design, participation rates, conceptualization of study variables, and demographic realities of secondary computer science education. With a descriptive study such as this one, the immediate limitations we can point to are due to selection bias (Remler & Van Ryzin, 2011) . In regards to self - selection bias, students have some control over which courses they enroll in in high school, with computer science often being an optional course not required for graduation (Zinth, 2016) . This leads to a bias towards students who already have an interest in computer science, or have been encouraged to do so, and thus may already have a higher sense of self - efficacy, belongingness, and persistence in computer science than the overall student population. The method of recruitment of students via their teach er introduced additional selection bias in gaining student participants. The expectation of teachers to conduct recruitment and collect study forms may have led to a lower participation rate and often resulted in entire classes of students not being recru ited due to teacher participants deciding to withdraw. There are also issues with bias as students can be directed to courses by their counselors and teachers, who may have pre - existing notions of which students are best suited for computer science (Margolis, 2008) . Finally, the decision of the students whether or not to participate in the study introduced yet another level of selection bias. Some of the selection bias was partially addres sed through the use of covariates such as pre - existing levels of self - efficacy, belongingness, and persistence, 71 Another limitation of this study would be that the participants were overwhelmingly identifie d as white or Asian and majority male. This limits the generalizations we can make from the results, as they do not fully reflect the experiences of students of color and intersections of other underrepresented groups of students (Vitores & Gil - Juárez, 2015) . Particularly troubling is the underrepresentation of f emale students of color in this sample, which mirrors a national problem within computer science . This could be addressed by recruiting students from multiple ethnicities and racial groups , resulting in large enough subsets of data for historically margin alized populations. Interviewing students about their decision to enroll in APCS, and any intersections they found between their experiences in APCS and their gender, race, or ethnicity, could also provide insight. Presenting the range of experiences stu dents describe across multiple settings, and finding similarities in such, would offer a more complete picture of APCS course experiences and speak to possible generalizable conclusions. The limitations of self - reported data are present in this study, with the possibility that students are biased to provide answers they believe socially acceptable, skewed toward more recent experiences, or may have a limited interpretation of the survey questions. Any of these biases can threaten the validity of the measur es and in some may play a role in the high mean scores for self - efficacy, belongingness, and persistence scores. One manner of addressing this could be through also complementing these self - reported measures with external measures that indicate similar co nstructs. For instance, while we do not have a method of externally measuring could frequency of interaction or voicing of sense of belonging or feeling isolated. External measures such as these are imperfect as well, but when coupled with the self - report data could strengthen the validity of the data. 72 Another limitation I attempted to address in this study was the conceptualization of gender as dichotomous (Glasser & Smith, 2008) . Gender was chosen as a variable rather than sex, due to the American Psychological Association ( APA ) guidance of gender being related to social groups , rather than sex which is often interpreted as being related to biological characteristics of students (2010) . I was mostly interest ed in how the courses offered would appeal to students in their social interactions within their classrooms , having no relation to biological designations. I recognized that the APA guidance is imperfect (Glasser & Smith, 2008) in that researchers are still often conflating sex and gender, and are defaulting to a dichotomou s definition of both. I attempted to address the range of possible gender identities by giving students a choice beyond male or female text desired and an option to not provide gender. The same choices we re made available for teachers when they reported the overall class gender percentages. However, I received only male or female responses from students, except for some invalid student responses that were very likely reactionary in nature toward the non - b inary gender option. While I had hoped this option offered space for non - binary and gender nonconforming students to self - identify for the study, I did not receive any such responses. In retrospect, the options I offered for gender designation of were no t affirming enough to encourage students to fully share their information. A more appropriate solution for future research would be a more explicit inclusion of multiple gender identities, rather than trying to capture these under a vague and possibly deh umanizing category I also see structural problems where study design may have influenced student responses, as an impersonal survey where the privacy of the data is unknown, may not be t he setting for a student to disclose this information. 73 In addition , there is a possible problem with the teacher reported class gender ratios, as some students may not yet be publicly identifying as thei r gender or may not match the school records of student gender that teachers likely used as a data source. Similar to race and ethnicity, even if I had students identifying as non - binary or gender nonconforming, the sample needed for quantitative analysis would have to be substantial. This, along with the personal nature of students identifying their gender, lends itself toward more qualitative methods in which trust can be established over time and in which individual, detailed accounts can offer more a within APCS. 5.3 Conclusion Advanced Placement Computer Science courses are but one context in which we can examine our efforts to address gender inequity within computer science education. T his study showed no - efficacy, belongingness, and persistence based solely up The findings suggest that female students encounter ed a slight decrease in their computer science self - efficacy in these course s, which does bolster the case for a more detailed examination of how these courses are taught and how female students describe their classroom experiences. The results of this study should not be interpreted to say that these APCS courses are of no benef it to the students, only that we should be careful not to automatically predict increases in female enrollment in undergraduate CS majors based solely on the increases we see in APCS courses . Addressing gender inequity is not solely about these the outcom e variables chosen for this study and more work needs to be done to use measures sensitive enough to capture nuances of student experience in computer science. The results from this study show that enrollment in courses designated as APCS - A or APCS - P does not predict 74 any differences in - efficacy, belongingness, or likelihood of persisting within computer science. In summary, results suggest that more work needs to be done and we cannot just put our faith in one course as the answer to address inequities within CS. 75 APPENDICES 76 APPENDIX A Survey for Students NOTE: Items that are in only one of the surveys are denoted as such 1.) First Name: 2.) Last Name: 3.) (PRE - TEST ONLY) Gender: a. Female b. Male c. Prefer not to say d. Other :____________ 4.) (PRE - TEST ONLY) Are you Spanish, Hispanic, Latino or none of these? 5.) (PRE - TEST ONLY) Choose one or more races that you consider yourself to be: a. White b. Black or African American c. Native American or Alaska Native d. Asian e. Native Hawaiian or Pacific Island f. Other:____________ 6.) Date of Birth: 7.) (PRE - TEST ONLY) What is your current grade level in school? a. Grade 9 b. Grade 10 c. Grade 11 d. Grade 12 e. Other: ____________ 8.) School Name: 9.) School City: 10.) School State: 77 11.) Teacher Last Name: Academic History 12.) Choose the statement that best describes your grades in high school up until now: a. b. c. d. e. f. Does not apply to me 13.) Choose the statement that best describes your grades in SCIENCE courses in high school up until now: a. b. c. d. e. f. Does not apply to me 14.) Choose the statement that best describes your grades in MATH courses in high school up until now: a. b. c. d. e. f. Does not appl y to me 15.) Which course(s) are you currently taking? a. AP Computer Science A only b. AP Computer Science Principles only c. Both AP Computer Science A and AP Computer Science Principles d. Neither AP Computer Science course (PRE - TEST ONLY) Previous computer science co urses (only completed by those answering A for question 15) 16.) (PRE - TEST ONLY) Have you previously taken the AP Computer Science Principles course? Y/N 17.) (PRE - TEST ONLY) Have you taken any non - AP Computer science courses 78 (PRE - TEST ONLY) Previous computer science courses (only completed by those answering B for question 15) 16.) (PRE - TEST ONLY) Have you previously taken the AP Computer Science A course? Y/N 17.) (PRE - TEST ONLY) Have you taken a non - AP computer science c ourses before this Belongingness This next set of questions deals with your feeling of belonging in this course. Today we have some questions we would like you to answer about your experience in your computer science courses a nd the computer science community. When we mention the computer science community, we are referring to the broad group of people involved in that field, including the students in a computer science course. We would like you to consider your membership in t he computer science community. By virtue of taking a computer science course, you could consider yourself a member of the computer science community. Given this broad definition of belonging to the computer science community, please respond to the followin g statements based on how you feel about that group and your membership in it. There are no right or wrong answers to any of these statements; we are interested in your honest reactions and opinions. Please read each statement carefully and indicate the nu mber that reflects your degree of agreement. Level of Agreement 1 Strongly agree 1.) I feel that I belong to the computer science community. 1 2 3 4 5 6 7 2.) I consider myself a member of the computer science world. 1 2 3 4 5 6 7 3.) I feel like I am part of the computer science community. 1 2 3 4 5 6 7 4.) I feel a connection with the computer s cience community. 1 2 3 4 5 6 7 5.) I feel like an outsider. 1 2 3 4 5 6 7 79 6.) I feel accepted. 1 2 3 4 5 6 7 7.) I feel respected. 1 2 3 4 5 6 7 8.) I feel disregarded. 1 2 3 4 5 6 7 9.) I feel valued. 1 2 3 4 5 6 7 10.) I feel neglected. 1 2 3 4 5 6 7 11.) I feel apprec iated. 1 2 3 4 5 6 7 12.) I feel excluded. 1 2 3 4 5 6 7 13.) I feel like I fit in. 1 2 3 4 5 6 7 14.) I feel insignificant. 1 2 3 4 5 6 7 15.) I feel at ease. 1 2 3 4 5 6 7 16.) I feel anxious. 1 2 3 4 5 6 7 17.) I feel comfortable. 1 2 3 4 5 6 7 18.) I feel tense. 1 2 3 4 5 6 7 19.) I feel nervous. 1 2 3 4 5 6 7 20.) I feel content. 1 2 3 4 5 6 7 21.) I feel calm. 80 1 2 3 4 5 6 7 22.) I feel inadequate. 1 2 3 4 5 6 7 23.) I wish I could fade into the background and not be noticed. 1 2 3 4 5 6 7 24.) I try to say as little as possible. 1 2 3 4 5 6 7 25.) I enjoy being an active participant. 1 2 3 4 5 6 7 26.) I wish I were invisible. 1 2 3 4 5 6 7 27.) I trust the testing materials to be unbiased. 1 2 3 4 5 6 7 28.) I have trust that I do not have to constantly prove myself. 1 2 3 4 5 6 7 29.) I trust my instruc tors to be committed to helping me learn. 1 2 3 4 5 6 7 30.) Even when I do poorly, I trust my instructors to have faith in my potential. 1 2 3 4 5 6 7 This next set of questions deals with how you believe you will perform in this course. There are no righ t or wrong answers to any of these statements; we are interested in your honest reactions and opinions. Please read each statement carefully, and indicate the number that reflects your degree of agreement Level of Agreement 1 Very true of me 31.) I believe I will receive an excellent grade in this class. 1 2 3 4 5 6 7 32.) for this course. 1 2 3 4 5 6 7 33.) taught in this course. 81 1 2 3 4 5 6 7 34.) instructor in this course. 1 2 3 4 5 6 7 35.) 1 2 3 4 5 6 7 36.) I expect to do well in this class. 1 2 3 4 5 6 7 37.) 1 2 3 4 5 6 7 38.) Considering the difficulty of this course, the teacher, and my skills, I think I will do well in this course. 1 2 3 4 5 6 7 Thi s next set of questions deals with how you may use your computer science knowledge in the future. There are no right or wrong answers to any of these statements; we are interested in your honest reactions and opinions. Please read each statement carefully, and indicate the number that reflects your degree of agreement Level of Agreement: 1 Strongly agree 39.) Knowledge of computer science will allow me to secure a better job 1 2 3 4 5 6 7 40.) My career goals do not require that I learn computer science skills 1 2 3 4 5 6 7 41.) I expect that learning computer science skills will help me to achieve my career goals 1 2 3 4 5 6 7 82 42.) I hope that my future career will require the use of computer science concepts 1 2 3 4 5 6 7 43.) Having background knowledge and understanding of computer science is valuable in and of itself 1 2 3 4 5 6 7 44.) I am interested in a career as a computer scientist 1 2 3 4 5 6 7 45.) I am interested in a career where knowledge of computer science would be useful. 1 2 3 4 5 6 7 46.) I plan to pursue a career that requires computer science skills 1 2 3 4 5 6 7 47.) I am interested in taking computer science courses in college. 1 2 3 4 5 6 7 48.) I am likely to teach myself computer science skills on my own. 1 2 3 4 5 6 7 49.) If given the opportunity, I would take more computer science courses 1 2 3 4 5 6 7 83 APPENDIX B Survey for Computer Science Teachers Thank you for volunteering to take part in this study. We expect to distribute these surveys at the beginning and end of the fall semester for all students and teachers. Your participation and gness and self - efficacy within computer science, along with their desire to persist in the field. 1.) First Name: 2.) Last Name: 3.) (PRE - TEST ONLY) Gender: a. Female b. Male c. Prefer not to say d. Other: ___________ 4.) (PRE - TEST ONLY) Are you Spanish, Hispanic, or Latino or n one of these? a. Yes b. No 5.) (PRE - TEST ONLY) Choose one or more races that you consider yourself to be: a. White b. Asian c. Black or African American d. Native Hawaiian or Pacific Islander e. Native American or Alaska Native f. Other:____________ 6.) (PRE - TEST ONLY) Date of Birth: 7.) School Name: 8.) School City 9.) School State 84 10.) I am currently teaching the course(s): a. AP Computer Science A b. AP Computer Science Principles 11.) My student in this (these) courses participated in the study: a. AP Computer Science A b. AP Computer Science Principles 12.) (PR E - TEST ONLY) What other computer science courses have you taught? 13.) (PRE - TEST ONLY) How many years, including the current academic year, have you taught AP Computer Science A 14.) (PRE - TEST ONLY) How many years, including the current academic year, have you t aught AP Computer Science Principles? 15.) (PRE - TEST ONLY) How many years, including the current academic year have you taught computer science courses? 16.) (PRE - TEST ONLY) How many years have you taught in general? 17.) (PRE - TEST ONLY) How many students of the fo llowing genders are currently enrolled in this course (if multiple classes/sections, please denote student for each class/section)? a. Male:____________ b. Female:___________ c. Other:_____________ 18.) (PRE - TEST ONLY) Which computer science courses are available for students to take at your school? a. Advanced Placement Computer Science A b. Advanced Placement Computer Science Principle c. Other:_____________________ 19.) (POST - TEST ONLY) If any students participating in the study withdrew from the class, please list their name(s ): ________________________________ 20.) (POST - TEST ONLY) If any students withdrew from the course, what reasons did the student(s) state for doing so? If multiple students listed, please state each with _____________________________________ ________________________________ 85 REFERENCES 86 REFERENCES American Psychological Association. (2010). Publication manual of the American Psychological Association. (6th ed.). Washi ngton, DC: American Psychological Association. Anderson, N. (2018, January 8). Expansion of AP computer science courses draws more girls and minorities. Washington Post . Retrieved from https://www.washingtonpost.com/local/education/expansion - of - ap - compute r - science - draws - more - girls - and - minorities/2018/01/08/cd5932d8 - e040 - 11e7 - 89e8 - edec16379010_story.html Aspray, W. (2016). Organizations That Help Women to Build Computing Careers. In Women and Underrepresented Minorities in Computing (pp. 165 203). Springer , Cham. Retrieved from https://link - springer - com.proxy1.cl.msu.edu/chapter/10.1007/978 - 3 - 319 - 24811 - 0_8 Bandura, A. (1977). Self - efficacy: Toward a Unifying Theory of Behavioral Change. Psychological Review , 84 (2), 191. Bandura, A. (1986). The Explanatory and Predictive Scope of Self - Efficacy Theory. Journal of Social and Clinical Psychology; New York , 4 (3), 359 373. https://doi.org/http://dx.doi.org.proxy2.cl.msu.edu/10.1521/jscp.1986.4.3.359 Bandura, A. (1997). Self - efficacy: the exercise of control . Ne w York: W.H. Freeman. Bandura, A. (2002). Social cognitive theory in cultural context. Applied Psychology , 51 (2), 269 290. Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivatio n. Psychological Bulletin , 117 (3), 497 529. https://doi.org/10.1037/0033 - 2909.117.3.497 Benakli, N., Kostadinov, B., Satyanarayana, A., & Singh, S. (2017). Introducing computational thinking through hands - on projects using R with applications to calculus, probability and data analysis. International Journal of Mathematical Education in Science and Technology , 48 (3), 393 427. https://doi.org/10.1080/0020739X.2016.1254296 Beyer, S. (2014). Why are women underrepresented in Computer Science? Gender differenc es in stereotypes, self - efficacy, values, and interests and predictors of future CS course - taking and grades. Computer Science Education , 24 (2 3), 153 192. https://doi.org/10.1080/08993408.2014.963363 Beyer, S., Rynes, K., Perrault, J., Hay, K., & Haller, S. (2003). Gender differences in computer science students (p. 49). ACM Press. https://doi.org/10.1145/611892.611930 87 Blaney, J. M., & Stout, J. G. (2017). Examining the Relationship Between Introductory Computing Course Experiences, Self - Efficacy, and Bel onging Among First - Generation College Women. In Proceedings of the 2017 ACM SIGCSE Technical Symposium on Computer Science Education (pp. 69 74). New York, NY, USA: ACM. https://doi.org/10.1145/3017680.3017751 Blikstein, P. (2013). Digital Fabrication and Invention. FabLabs: Of Machines, Makers and Inventors , 1 21. Brown, B. (2015). Project and Problem Based Learning: PLTW and AP CS Principles. Acm , 52. Chemers, M. M., Zurbriggen, E. L., Syed, M., Goza, B. K. , & Bearman, S. (2011). The Role of Efficacy and Identity in Science Career Commitment Among Underrepresented Minority Students: Efficacy and Identity in Science Career Commitment. Journal of Social Issues , 67 (3), 469 491. https://doi.org/10.1111/j.1540 - 45 60.2011.01710.x Statistical Analysis Report. NCES 2014 - 001. National Center for Education Statistics . Retrieved from http://eric.ed.gov/?id=ED544470 Cheryan, S., Plaut, V. C., Davies, P. G., & Steele, C. M. (2009). Ambient belonging: How stereotypical cues impact gender participation in computer science. Journal of Personality and Social Psychology , 97 (6), 1045 1060. https://doi.org/10.1037/a0016239 Code.org. (2017, Marc h 27). University computer science finally surpasses its 2003 peak! Retrieved April 23, 2017, from https://medium.com/anybody - can - learn/university - computer - science - finally - surpasses - its - 2003 - peak - ecefa4c8d77d Cohoon, J. M., & Lord, H. (2006). A faculty ro M. Trauth (Ed.), Encyclopedia of gender and information technology (pp. 297 303). Hershey, PA: Idea Group Reference. College Board. (2017). AP Program Participation and Performance Data 2017: Program Summary Report 2017 . Retrieved from https://secure - media.collegeboard.org/digitalServices/pdf/research/2017/Program - Summary - Report - 2017.pdf Cooper, K. M., Krieg, A., & Brownell, S. E. (2018). Who perceives they are smarter? Exploring the influence of student cha racteristics on student academic self - concept in physiology. Advances in Physiology Education , 42 (2), 200 208. https://doi.org/10.1152/advan.00085.2017 Creswell, J. W. (2008). Educational Research: Planning, Conducting, and Evaluating Quantitative and Qua litative Research (3 edition). Upper Saddle River, N.J: Prentice Hall. 88 Crowell, R. (2016). Gender Bias and the Evaluation of Players: Voice and Gender in Narrated Gameplay Videos. Dissertations and Theses . https://doi.org/10.15760/etd.3150 CS Education R esearch Group. (2014). Computer Science Unplugged. Retrieved January 12, 2014, from http://csunplugged.org/ Cundiff, J. L., Vescio, T. K., Loken, E., & Lo, L. (2013). Do gender - science stereotypes predict science identification and science career aspirati ons among undergraduate science majors? , 16 (4), 541 554. https://doi.org/http://dx.doi.org.proxy2.cl.msu.edu/10.1007/s11218 - 013 - 9232 - 8 the self - determination of behavior. Psychological Inquiry , 11 (4), 227 268. and the Self - Determination of Behavior. Psychological Inquiry , 11 (4), 227 268. Dollar, D., & Gatti, R. (1999). Gender inequality, income, and growth: are good times good for women? (Vol. 1). Development Research Group, The World Bank Washington, DC. Retrieved from http://darp.lse.ac.uk/frankweb/courses/EC50 1/DG.pdf Dougherty, D. (2012). The maker movement. Innovations , 7 (3), 11 14. DuBow, W., Kaminsky, A., & Weidler - Lewis, J. (2017). Multiple Factors Converge to Influence Computing in Science & Engi neering , 19 (3), 30 39. https://doi.org/10.1109/MCSE.2017.42 Eccles, J. S., Barber, B., & Jozefowicz, D. (1999). Linking gender to educational, occupational, and recreational choices: Applying the Eccles et al. model of achievement - related choices. In Sexi sm and stereotypes in modern society: The gender science of Janet Taylor Spence (pp. 153 192). Washington, DC, US: American Psychological Association. Ericson, B. (2014, January 1). Detailed AP CS 2013 Results: Unfortunately, much the same. Retrieved Jan uary 12, 2017, from https://computinged.wordpress.com/2014/01/01/detailed - ap - cs - 2013 - results - unfortunately - much - the - same/ Espinoza, P., Arêas Da Luz Fontes, A. B., & Arms - chavez, C. J. (2014). Attributional gender Social , 17 (1), 105 126. https://doi.org/http://dx.doi.org.proxy1.cl.msu.edu.proxy2.cl.msu.edu/10.1007/s11218 - 013 - 9226 - 6 Field, A. P., Miles, J., & Field, Z. (2012). Discovering statistics using R Oaks, Calif: Sage. 89 Finch, W. H. (2014). Multilevel modeling using R . Boca Raton, FL: CRC Press, Taylor & Francis Group. Frucgt, V. G., & Cook, G. L. (1994). Further research on the accuracy of s - reported grade point averages, SAT scores, and course grades. Perceptual and Motor Skills , 79 (2), 743 746. Research: The Problem, Its Sources, and Recommend ations for Practice. Educational Researcher , 37 (6), 343 350. https://doi.org/10.3102/0013189X08323718 Good, C., Rattan, A., & Dweck, C. S. (2012). Why do women opt out? Sense of belonging and Journal of Personality a nd Social Psychology; Washington , 102 (4), 700. Goode, J. (2008). Increasing diversity in k - 12 computer science: strategies from the field. In ACM SIGCSE Bulletin (Vol. 40, pp. 362 366). ACM. Retrieved from http://dl.acm.org/citation.cfm?id=1352259 Goode, J., Margolis, J., & Chapman, G. (2014). Curriculum is not enough: the educational theory and research foundation of the exploring computer science professional development model (pp. 493 498). ACM Press. https://doi.org/10.1145/2538862.2538948 Goodenow, C. (1993). Classroom Belonging among Early Adolescent Students: Relationships to Motivation and Achievement. The Journal of Early Adolescence , 13 (1), 21 43. https://doi.org/10.1177/0272431693013001002 Google Inc. & Gallup Inc. (2014). Women Who Choose CS: What Really Matters. Google Inc. & Gallup Inc. Retrieved from https://goo.gl/rLX6ax Google Inc. & Gallup Inc. (2015). Images of Computer Science: Perceptions Among Students, Parents and Educators in the U.S. Google Inc. & Gallup Inc. Retrieved from https ://goo.gl/F3SSWH Grover, S., Pea, R., & Cooper, S. (2014). Remedying Misperceptions of Computer Science Among Middle School Students. In Proceedings of the 45th ACM Technical Symposium on Computer Science Education (pp. 343 348). New York, NY, USA: ACM. h ttps://doi.org/10.1145/2538862.2538934 Halverson, E. R., & Sheridan, K. M. (2014). The Maker Movement in Education. Harvard Educational Review , 84 (4), 495 504. Hambrusch, S., Hoffmann, C., Korb, J. T., Haugan, M., & Hosking, A. L. (2009). A Multidiscipli nary Approach Towards Computational Thinking for Science Majors. ACM SIGCSE Bulletin , 41 (1), 183 187. 90 Hamilton, M., Luxton - Wong, E. (2016). Gender Equity in Computing: International Faculty Perceptions and Current Practices. In Proceedings of the 2016 ITiCSE Working Group Reports (pp. 81 102). New York, NY, USA: ACM. https://doi.org/10.1145/3024906.3024911 Hausmann, L. R. M., Schofield, J. W., & Woods, R. L. (2007). Sense of Belongi ng as a Predictor of Intentions to Persist Among African American and White First - Year College Students. Research in Higher Education , 48 (7), 803 839. https://doi.org/10.1007/s11162 - 007 - 9052 - 9 Ingels, S. J. (1990). National Education Longitudinal Study of 1988 (NELS:88) - Student . Retrieved from https://nces.ed.gov/surveys/nels88/ Inzlicht, M., & Ben - Zeev, T. (2000). A Threatening Intellectual Environment: Why Females Are Susceptible to Experiencing Problem - Solving Defici ts in the Presence of Males. Psychological Science , 11 (5), 365 371. https://doi.org/10.1111/1467 - 9280.00272 Inzlicht, M., & Good, C. (2005). How environments can threaten academic performance, self - knowledge, and sense of belonging. In S. Levin & C. V. La ar (Eds.), Stigma and Group Inequality: Social Psychological Perspectives . Mahwah, NJ: Psychology Press. Kafai, Y. B., & Burke, Q. (2014). Beyond Game Design for Broadening Participation: Building New Clubhouses of Computing for Girls. In Proceedings of G ender and IT Appropriation. Science and Practice on Dialogue - Forum for Interdisciplinary Exchange (p. 21). Siegen, Germany: European Society for Socially Embedded Technologies. Retrieved from http://dl.acm.org/citation.cfm?id=2670301 Kafai, Y. B., & Bur ke, Q. (2015). Constructionist Gaming: Understanding the Benefits of Making Games for Learning. Educational Psychologist , 50 (4), 313 334. https://doi.org/10.1080/00461520.2015.1124022 Kafai, Y. B., Fields, D., & Searle, K. (2014). Electronic Textiles as D isruptive Designs: Supporting and Challenging Maker Activities in Schools. Harvard Educational Review , 84 (4), 532 556. Kafai, Y. B., Lee, E., Searle, K., Fields, D., Kaplan, E., & Lui, D. (2014). A Crafts - Oriented Approach to Computing in High School: Int roducing Computational Concepts, Practices, and Perspectives with Electronic Textiles. Trans. Comput. Educ. , 14 (1), 1:1 1:20. https://doi.org/10.1145/2576874 Kafai, Y. B., Peppler, K. A., & Chapman, R. N. (2009). The Computer Clubhouse: Constructionism an d Creativity in Youth Communities (50045th edition). New York: Teachers College Press. 91 Kreft, I., & Leeuw, J. de. (1998). Introducing multilevel modeling Calif: Sage. Kuncel, N. R., Credé, M., & Thomas, L. L. (2005). The Validity of Self - Reported Grade Point Averages, Class Ranks, and Test Scores: A Meta - Analysis and Review of the Literature. Review of Educational Research; Washington , 75 (1), 63 82. Lee, A. (2015). Determining the effects of computer science education at the seco ndary level on STEM major choices in postsecondary institutions in the United States. Computers & Education , 88 , 241 255. https://doi.org/10.1016/j.compedu.2015.04.019 Lee, I., Martin, F., & Apone, K. (2014). Integrating Computational Thinking Across the K 8 Curriculum. ACM Inroads , 5 (4), 64 71. https://doi.org/10.1145/2684721.2684736 Lent, R. W., Brown, D., & Hackett, G. (2002). Social Cognitive Career Theory. In Career choice and development (4th ed, pp. 255 311). San Francisco, CA: Jossey - Bass. Lent, R. W., Brown, S. D., & Hackett, G. (1994). Toward a Unifying Social Cognitive Theory of Career and Academic Interest, Choice, and Performance. Journal of Vocational Behavior , 45 (1), 79 122. https://doi.org/10.1006/jvbe.1994.1027 Lent, R. W., Brown, S. D., & Larkin, K. C. (1986). Self - Efficacy in the Prediction of Academic Performance and Perceived Career Options. Journal of Counseling Psychology; Washington, D.C. , 33 (3), 265. tting in 32). ACM Press. https://doi.org/10.1145/2960310.2960332 Lewis, K. L., Stout, J. G., Pollock, S. J., Finkelstein, N. D., & Ito, T. A. (2016). Fitting in or opting out: A review of key social - psychological factors influencing a sense of belonging for women in physics. Physical Review Physics Education Research , 12 (2). https://doi.org/10.1103/PhysRevPhysEducRes.12.020110 Lishinski, A. (2017). Cognitive, Affective, and Dispositional Components o f Learning to Program (Doctoral Dissertation). Michigan State University. Lishinski, A., Yadav, A., Good, J., & Enbody, R. (2016). Learning to Program: Gender - Efficacy on Perform ance. In Proceedings of the 2016 ACM Conference on International Computing Education Research (pp. 211 220). New York, NY, USA: ACM. https://doi.org/10.1145/2960310.2960329 Maas, C. J. M., & Hox, J. J. (2004). The influence of violations of assumptions on multilevel parameter estimates and their standard errors. Computational Statistics & Data Analysis , 46 (3), 427 440. https://doi.org/10.1016/j.csda.2003.08.006 92 Margolis, J. (2008). Stuck in the shallow end: education, race, and computing . Cambridge, Mass: MIT Press. Margolis, J. (2013). Unlocking the clubhouse: a decade later and now what? In Proceeding of the 44th ACM technical symposium on Computer science education (pp. 9 10). ACM. Retrieved from http://dl.acm.org/citation.cfm?id=2445202 Margolis, J., & Fisher, A. (2003). Unlocking the Clubhouse: Women in Computing . Cambridge, Mass.: The MIT Press. Margolis, J., Fisher, A., & Miller, F. (2000). The anatomy of interest: Women in undergraduate computer science. Y. , 28 (1), 104 105. Performance. Personality and Social Psychology Bulletin , 28 (9), 1183 1193. https://doi.org/10.1177/01461672022812004 Master, A., Cheryan, S., & Meltz off, A. N. (2016). Computing whether she belongs: Stereotypes Journal of Educational Psychology , 108 (3), 424 437. https://doi.org/10.1037/edu0000061 McInerney, C. R., DiDonato, N. C., G CAREERS: A QUALITATIVE STUDY. Information Technology, Learning, and Performance Journal , 24 (2), 35. Murphy, M. C., Steele, C. M., & Gross, J. J. (2007). Signali ng Threat: How Situational Cues Affect Women in Math, Science, and Engineering Settings. Psychological Science , 18 (10), 879 885. https://doi.org/10.1111/j.1467 - 9280.2007.01995.x National Science Board. (2016). Science and Engineering Indicators 2016. Retr ieved from https://www.nsf.gov/statistics/2016/nsb20161/#/ NCES. (1988). National Education Longitudinal Study of 1988 (NELS:88) - Overview. Retrieved June 22, 2018, from https://nces.ed.gov/surveys/nels88/ Niehaus, E., Campbell, C. M., & Inkelas, K. K. (2014). HLM Behind the Curtain: Unveiling Decisions Behind the Use and Interpretation of HLM in Higher Education Research. Research in Higher Education , 55 (1), 101 122. Obama, B. H. (2016, January 12). Remarks of President Barack Obama State of the Unio n Address As Delivered. Retrieved January 14, 2016, from https://www.whitehouse.gov/the - press - office/2016/01/12/remarks - president - barack - obama - %E2%80%93 - prepared - delivery - state - union - address 93 Oudshoorn, N., Rommes, E., & Stienstra, M. (2004). Configuring t he User as Everybody: Gender and Design Cultures in Information and Communication Technologies. Science, Technology, & Human Values , 29 (1), 30 63. https://doi.org/10.1177/0162243903259190 Pajares, F., & Miller, M. D. (1994). Role of Self - Efficacy and Self - Concept Beliefs in Mathematical Problem Solving: A Path Analysis. Journal of Educational Psychology , 86 (2), 193. Pajares, F., & Valiante, G. (1997). Influence of Self - The Journal of Educational Research , 90 (6), 3 53 360. https://doi.org/10.1080/00220671.1997.10544593 Partovi, H. (2013). With support from Obama to Shakira, Apple to Zuckerberg, the Hour of Code is Here. Retrieved March 24, 2014, from http://codeorg.tumblr.com/post/69469632752/hourofcodeishere Picke tt, C. L., Gardner, W. L., & Knowles, M. (2004). Getting a Cue: The Need to Belong and Enhanced Sensitivity to Social Cues. Personality and Social Psychology Bulletin , 30 (9), 1095 1107. https://doi.org/10.1177/0146167203262085 Pintrich, P. R., Smith, D. A . F., Garcia, T., & McKeachie, W. J. (1991). A Manual for the Use of the Motivated Strategies for Learning Questionnaire (MSLQ) . Retrieved from https://www.researchgate.net/publication/271429287_A_Manual_for_the_Use_of_the_M otivated_Strategies_for_Lea rning_Questionnaire_MSLQ Pittman, L. D., & Richmond, A. (2007). Academic and Psychological Functioning in Late Adolescence: The Importance of School Belonging. The Journal of Experimental Education , 75 (4), 270 290. https://doi.org/10.3200/JEXE.75.4.270 - 29 2 Plantenga, J. (2015). The economics of gender equality; a review of the literature in three propositions and two questions . WWWforEurope Working Paper. Retrieved from https://www.econstor.eu/handle/10419/125759 Remler, D. K., & Van Ryzin, G. G. (2011). Research Methods in Practice: Strategies for Description and Causation . Thousand Oaks, Calif: SAGE Publications. Riegel - Gen der and Society , 26 (2), 290 322. Rommes, E., Oost, E. V., & Oudshoorn, N. (1999). Gender in the Design of the Digital City of Amsterdam. Information, Communication & Society , 2 (4), 476 495. https://doi.org/10.1080/136911899359510 Ryoo, J., Goode, J., & M argolis, J. (2015). It takes a village: supporting inquiry - and equity - oriented computer science pedagogy through a professional learning community. 94 Computer Science Education , 25 (4), 351 370. https://doi.org/10.1080/08993408.2015.1130952 Searle, K. A., F ields, D. A., Lui, D. A., & Kafai, Y. B. (2014). Diversifying high school 82). ACM Press. https://doi.org/10.1145/2632320.2632352 Searle, K. A., & Kafai, Y. B. (2015). Culturally Responsive Making with American Indian Girls: Bridging the Identity Gap in Crafting and Computing with Electronic Textiles (pp. 9 16). ACM Press. https://doi.org/10.1145/2807565.2807707 Shapiro, M., Grossman, D., Carter, S., Martin, K., Deyton, P., & Hammer, D. (201 5). Middle Middle School Journal; Columbus , 46 (5), 3 13. Sheridan, K. M., Halverson, E. R., Litts, B. K., Brahms, L., Jacobs - Priebe, L., & Owens, T. (2014). Learning in the making: A comparative case st udy of three makerspaces. Harvard Educational Review , 84 (4), 505 531. Shumow, L., & Schmidt, J. A. (2013). Academic grades and motivation in high school science belief s, and practices. In R. Haumann & G. Zimmer (Eds.), Handbook of academic performance: predictors, learning strategies and influences of gender (pp. 53 72). Smith, J. L., Brown, E. R., Thoman, D. B., & Deemer, E. D. (2015). Losing its expected scientists. , 18 (3), 443 466. https://doi.org/http://dx.doi.org.proxy1.cl.m su.edu.proxy2.cl.msu.edu/10.1007/s11218 - 015 - 9296 - 8 - Dominated STEM Fields As a Function of Effort Expenditure Concerns. Personality and Social Psychology Bulletin , 39 (2), 131 143. https://doi.org/10.1177/0146167212468332 Steele, C. M. (1997). A threat in the air: How stereotypes shape intellectual identity and performance. American Psychologist , 52 (6), 613. Stee le, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology , 69 (5), 797. Tellhed, U., Bäckström, M., & Björklund, F. (2017). Will I Fit in and Do Well? The Imp ortance of Social Belongingness and Self - Efficacy for Explaining Gender Differences in Interest in STEM and HEED Majors. Sex Roles , 77 (1 2), 86 96. https://doi.org/10.1007/s11199 - 95 016 - 0694 - y The College Board. (2014). AP Computer Science A Course Descripti on. The College Board. Retrieved from https://secure - media.collegeboard.org/digitalServices/pdf/ap/ap - computer - science - a - course - description.pdf The College Board. (2016a). AP Computer Science Principles Curriculum Framework. Retrieved from https://advance sinap.collegeboard.org/stem/computer - science - principles/course - details The College Board. (2016b). AP Program Participation and Performance Data 2016 - National Report. The College Board. Retrieved from https://research.collegeboard.org/programs/ap/data/p articipation/ap - 2016 Trump, D. J. (2017, September 25). Presidential Memorandum for the Secretary of Education. Retrieved October 16, 2017, from https://www.whitehouse.gov/the - press - office/2017/09/25/memorandum - secretary - education Turkle, S. (1997). Life on the Screen: Identity in the Age of the Internet . New York: Simon & Schuster. Turkle, S., & Papert, S. (1990). Epistemological Pluralism: Styles and Voices within the Computer Culture. Signs , 16 (1), 128 157. Turkle, S., & Papert, S. (1992). Epistemolo gical pluralism and the revaluation of the concrete. Journal of Mathematical Behavior , 11 (1), 3 33. U.S. Department of Education. (2017). WebCASPAR . Institute of Education Sciences, National Center for Science and Engineering Statistics. Retrieved from ht tps://ncsesdata.nsf.gov/webcaspar/ Vitores, A., & Gil - examination of the deployment of research on the gender gap in computer science. Journal of Gender Studies , 1 15. https://doi.org/10 .1080/09589236.2015.1087309 Voogt, J., Fisser, P., Good, J., Mishra, P., & Yadav, A. (2015). Computational thinking in compulsory education: Towards an agenda for research and practice. Education and Information Technologies , 20 (4), 715 728. https://doi.o rg/10.1007/s10639 - 015 - 9412 - 6 Wagh, A., Gravel, B., & Tucker - Raymond, E. (2017). The Role of Computational Thinking Practices in Making: How Beginning Youth Makers Encounter & Appropriate CT Practices in Making. In Proceedings of the 7th Annual Conference on Creativity and Fabrication in Education (p. 9:1 9:8). New York, NY, USA: ACM. https://doi.org/10.1145/3141798.3141808 96 Walton, G. M., Logel, C., Peach, J. M., Spencer, S. J., & Zanna, M. P. (2015). Two brief and achievement in engineering. Journal of Educational Psychology , 107 (2), 468 485. https://doi.org/10.1037/a0037461 Yadav, A., Gretter, S., & Good, J. (2017). Computer science for all: Role of gender in middl e American Educational Research Association, San Antonio, TX. Yadav, A., Hong, H., & Stephenson, C. (2016). Computational Thinking for All: Pedagogical Approaches to Embedd ing 21st Century Problem Solving in K - 12 Classrooms. TechTrends , 1 4. https://doi.org/10.1007/s11528 - 016 - 0087 - 7 Yadav, A., Mayfield, C., Zhou, N., Hambrusch, S., & Korb, J. T. (2014). Computational Thinking in Elementary and Secondary Teacher Education. A CM Transactions on Computing Education , 14 (1), 1 16. https://doi.org/10.1145/2576872 Zinth, J. (2016). Computer Science in High School Graduation Requirements. Education Commission of the States, Education Trends . Retrieved from http://eric.ed.gov/?id=ED5 68885 Zweben, S. (2013). Computing Degree and Enrollment Trends. Computing Research . Zweben, S., & Bizot, B. (2016). 2015 Taulbee Survey. Computing , 18 (5). Retrieved from http://craorg.c.presscdn.com/wp - content/uploads/2015/01/Taulbee - 2012 - 13.pdf