Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Field dependence-independence, logical-mathematical intelligence, and task persistence as predictors of engineering student’s performance in Chile
(USC Thesis Other)
Field dependence-independence, logical-mathematical intelligence, and task persistence as predictors of engineering student’s performance in Chile
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Running head: PREDICTING ENGINEERING STUDENT’S PERFORMANCE 1 FIELD DEPENDENCE-INDEPENDENCE, LOGICAL-MATHEMATICAL INTELLIGENCE, AND TASK PERSISTENCE AS PREDICTORS OF ENGINEERING STUDENT’S PERFORMANCE IN CHILE by Yael Stekel Schwarz A Dissertation Presented to the FACULTY OF THE USC ROSSIER SCHOOL OF EDUCATION UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF EDUCATION December 2016 Copyright 2016 Yael Stekel Schwarz PREDICTING ENGINEERING STUDENT’S PERFORMANCE 2 Acknowledgements I would like to, first and foremost, thank my parents; they have loved and supported me since my first day of life. I want to thank my dad, Moises Stekel (Z.L.), for allowing me to pursue my dreams and for protecting me in that journey. He was always there for me, for any reason and whenever I needed him, to give me his wise advice and help me in so many different ways. I want to thank my mom, Myriam Schwarz, for being my emotional support since ever. She, more than any other person, has unconditionally believed in me. I am especially thankful to my husband, Roni Gloger, and our two little girls, Sara (5 years old) and Noemi (3 years old). Without their support, even thinking about a doctorate would not have been possible. Roni has been my partner for the last 12 years, and we have built our dreams together. He sowed in me the inquietude and the motivation to move from Chile to the US in order to expand our horizons and learn from new experiences. I thank him for supporting my decision to pursue my master’s degree and then my Doctorate in Education at USC. In addition, I would like to thank Sara and Noemi, who, even though too young to participate in these family decisions, have had to spend longer hours in preschool and endure, at times, a tired mom to allow me concrete my educational projects. I would like to thank my committee professors: Dr. Dennis Hocevar, Dr. Miguel Torres and Dr. Patricia Tobey, all of whom actively participated along the process. I am especially grateful to Dr. Dennis Hocevar, my committee chairperson. His knowledge, dedication, support, and guidance were critically important in order to successfully complete my dissertation. I also offer my sincere gratitude to Dr. Miguel Torres, who is part of the school of engineering at the Pontificia Universidad Católica de Chile. Without his commitment and crucial participation during the data collection process, this research would not have been possible. In addition, I PREDICTING ENGINEERING STUDENT’S PERFORMANCE 3 would like to thank Dr. Robert Keim, who served as my dissertation chairperson during an important part of the time. Finally, I would like to thank all those professors and classmates who deeply influenced me along the way from the very first day of classes. Because of the profound admiration I have for these professors and the mark they have left on me, both inspirational and on my learning, I am especially grateful to Dr. Rousseau, Dr. Mendoza, Dr. Samkian and Dr. Polikoff. Thank you all! PREDICTING ENGINEERING STUDENT’S PERFORMANCE 4 Table of Contents Acknowledgements 2 List of Tables 6 List of Figures 8 Abstract 9 CHAPTER ONE: OVERVIEW OF THE STUDY 10 Background of the Problem 11 Statement of the Problem 15 Purpose of the Study 16 Significance of the Study 18 Limitations and Delimitations 18 Organization of the Study 20 CHAPTER TWO: LITERATURE REVIEW 21 Chilean Higher Education’s Institutions, Programs and Admission System 21 Pontificia Universidad Católica de Chile’s School of Engineering 22 The Talent and Inclusion Program at PUC 23 The PUC School of Engineering's Curriculum 24 First stage: Bachelor of Science in Engineering 24 Second Stage: Posterior Degree, Early Employment or Entrepreneurship 25 Field Dependence-Independence, Logical-Mathematical Intelligence, Task Persistence 26 Field Dependence-Independence 26 Logical-Mathematical Intelligence 28 Task Persistence 29 CHAPTER THREE: METHODOLOGY 31 Population and Sample 31 Instrumentation 32 Group Embedded Figures Test - Computerized Version (GEFT-C) 32 Test of Superior Logical Intelligence (TILS) 33 Numerical Ingenuity Test (TIN) 34 Data Collection 35 Analysis 36 CHAPTER FOUR: RESULTS 39 Descriptive Statistics 39 Dependent Variables 39 Independent Variables 44 Internal Consistency of the Tests 48 Field Dependence-Independence and Student Performance 50 The Relationship between Field Dependence-Independence and Student PerformanceError! Bookmark not defined. Field Dependence-Independence as a Predictor of Students Performance 50 Logical-Mathematical Intelligence and Students Performance 56 The Relationship Between Logical-Mathematical Intelligence and Student Performance 57 Logical-Mathematical Intelligence as a Predictor of Students Performance 57 Task Persistence and Students Performance 63 The Relationship Between Task Persistence and Students Performance 64 Task Persistence as a Predictor of Students Performance 68 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 5 CHAPTER FIVE: DISCUSSION 74 Discussion of Findings 74 Field Dependence-Independence as a Predictor of Academic Performance of Engineering Students During Their First Year of College 74 Logical-Mathematical Intelligence as a Predictor of Academic Performance of Engineering Students During Their First Year of College 75 Task Persistence as a Predictor of Academic Performance of Engineering Students During Their First Year of College 76 Discussion of Findings in Relation to Literature 79 Implications for Practice 80 Propositions for Future Research 81 Final Conclusions 82 References 84 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 6 List of Tables Table 1: Students in the Sample from the PUC School of Engineering 32 Table 2: Descriptive Data for Dependent Variables 41 Table 3: Spearman's rho Correlations Coefficients for DV’s 44 Table 4: Descriptive Data for Independent Variables 45 Table 5: Spearman's rho Correlations Coefficients for IV’s 48 Table 6: Cronbach’s Alpha for Sub-Dimensions in TIN 50 Table 7: Spearman's rho Correlations Coefficients for FDI (measured through the GEFT-C) and DV’s 51 Table 8: Group Statistics for Independent Samples Test Between Students Who Scored Above and Below the Mean in the GEFT-C (For Field Dependence-Independence) 52 Table 9: Independent Samples Test Between Students Who Scored Above and Below the Mean in the GEFT-C (For Field Dependence-Independence) 53 Table 10: Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in Calculus I 55 Table 11: Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in Linear Algebra 55 Table 12: Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st Semester GPA 56 Table 13: Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st year Cumulative GPA 56 Table 14: Spearman's rho Correlations Coefficients for Logical-Mathematical Intelligence (measured through the TILS) and DV’s 58 Table 15: Group Statistics for Independent Samples Test Between Students Who Scored Above and Below the Mean in the TILS (for Logical-mathematical Intelligence) 59 Table 16: Independent Samples Test for Students Who Scored Above and Below the Mean in the TILS (for Logical-mathematical Intelligence) 60 Table 17: Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in Calculus I 62 Table 18: Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in Linear Algebra 62 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 7 Table 19: Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in Statics and Dynamics 62 Table 20: Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st Semester GPA 63 Table 21: Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st Year Cumulative GPA 63 Table 22: Spearman's rho Correlations Coefficients for Task Persistence and DV’s 66 Table 23: Summary of Regression Analysis for Variables Predicting Student’s Performance in Calculus I 71 Table 24: Summary of Regression Analysis for Variables Predicting Student’s Performance in Chemistry II 72 Table 25: Summary of Regression Analysis for Variables Predicting Student’s Performance in Linear Algebra 72 Table 26: Summary of Regression Analysis for Variables Predicting Student’s Performance for 1 st Semester GPA 73 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 8 List of Figures Figure 1. Strength of the Relationship Between Performance and Socio-economic Background (PISA, 2009) 11 Figure 2. Distribution of Students enrolled in Private Paid, Private Subsidized and Municipalized Schools by Students’ Family Income 13 Figure 3. Students who Take the PSU and are Enrolled in a University by School’s Provider 14 Figure 4. Curriculum for the Bachelor’s Degree in Science of Engineering 26 Figure 5. Example of the Type of Problems in the GEFT Test 33 Figure 6. Example of the Type of Problems in the TILS Test 34 Figure 7. Example of the Type of Problems in the Numerical Ingenuity Test 35 Figure 8. Histograms for Dependent Variables: Mandatory Classes 1 st Semester 41 Figure 9. Histograms for Dependent Variables: Mandatory Classes 2 nd Semester 42 Figure 10. Histograms for Dependent Variables: Elective Classes 1 st Year 42 Figure 11. Histograms for Dependent Variables: Grade Point Average and Cumulative Grade Point Average 43 Figure 12. Histograms for Field Dependence Independence (GEFT-C) 46 Figure 13. Histogram for Logical-Mathematical Intelligence (TILS) 46 Figure 14. Histograms for Task Persistence (TIN) 47 Figure 15. Histogram for PSU Score 47 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 9 Abstract This study used statistical techniques to analyze field dependence-independence (FDI), logical- mathematical intelligence (LMI) and task persistence (TP) as possible predictors of engineering student’s performance in Chile. The purpose of this study was to identify additional measures to improve the selection criteria of students applying to the school of engineering at the Pontificia Universidad Católica de Chile via the Talent and Inclusion Program. The sample was composed of 65 students admitted via the Talent and Inclusion Program and 58 students admitted via regular admission, all of them entering engineering the 2015 school year. The 123 students were tested in FDI, LMI and TP using the GEFT-C (Group Embedded Figure Test – Computerized Version), the TILS (in Spanish, Test de Inteligencia Lógica Superior) and the TIN (in Spanish, Test de Ingenio Numérico), respectively. The data were analyzed utilizing different statistical techniques, first without controls and then controlling for PSU (in Spanish, Prueba de Selección Universitaria), the university’s current selection test. Findings from this study indicate that even though FDI and LMI can predict students’ performance in the most mathematics-intense courses and also GPA, their capacity to predict student performance is drastically smaller than PSU scores. In addition, and given some important problems of validity and reliability of the TIN, the results from this study were not conclusive as to whether TP might predict student grades. Given the broad literature that supports the relationship between TP and students’ performance (but uses self-reported questionnaires instead of assessments), it is strongly recommended that researchers pursue investigations to identify reliable and valid tests to assess TP that are not self- report and that can be used for selection purposes. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 10 CHAPTER ONE: OVERVIEW OF THE STUDY Historically, Chilean universities, especially more selective institutions, have strongly relied on the University Selection Test (hereafter PSU, which, in Spanish, stands for Prueba de Selección Universitaria) to choose their students. Because PSU scores are highly correlated with students’ socioeconomic background, this selection process perpetuates educational inequities that start at a very early age and continue through K-12 education. In this context, some Chilean universities have seen the limitations of the current admission system for attracting talented students from more diverse socioeconomic backgrounds and decided to implement programs that open a side door for admitting those students. Most of these programs allow students from families of the lowest income quintiles to be admitted if they meet certain prerequisites, such as ranking in the top 10% of their school cohort. The School of Engineering at the Pontificia Universidad Católica de Chile (PUC), which has the most selective engineering program in Chile, launched, in 2010, the Talent and Inclusion Program (hereafter, T+I Program). Five years after the beginning of the T+I Program, the school of engineering still investigates the best non-academic predictors of student success that could be used as measures for selecting engineering students who are admitted via this program. In this context, the following study evaluates field dependence-independence, logical intelligence, and task persistence as possible predictors of students’ performance in engineering. Even though this study is developed in the context of the T+I Program from the PUC School of Engineering, the findings of the study could be of interest to those in other STEM careers at the PUC and other universities in Chile. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 11 Background of the Problem Chile advocates for an equitable educational system where all students have the same learning opportunities regardless of their socioeconomic status (SES). Nevertheless, in Chile, as in many countries, the K-12 educational system has not been able to be a social equalizer and students’ SES strongly determines their educational outcomes. According to the Organisation for Economic Co-operation and Development’s (OECD) report on equity and quality in education (Figure 1), Chile is located as one of the more inequitable countries and has poorer Program for International Student Assessment (PISA) results (OECD, 2012). How to read this chart: This graph shows the extent to which student performance is determined by socioeconomic differences by plotting the average level of performance (y-axis) and the variance in performance explained by the socioeconomic background of students (x-axis). Source: OECD (2012). Figure 1. Strength of the Relationship Between Performance and Socioeconomic Background (PISA, 2009) PREDICTING ENGINEERING STUDENT’S PERFORMANCE 12 In addition to an overall deficient performance of Chilean students, the results on national standardized tests also reveal an educational system that is highly segregated both socially and academically. A study by Duarte, Bos, Moreno and Morduchowicz (2013) shows trends from 1999 to 2011 in socioeconomic and academic segregation in schools and the changes in the relationship of SES to academic achievement between and within schools as well as the evolution in learning gaps between students from different socioeconomic backgrounds. Their results suggest high social and academic segregation that, although declining, remains uneven among grades. In regards to social segregation, Duarte et. al. (2013) found that the probability that students of a particular SES attend the same school as students of a similar socioeconomic background is greater than 60%. That is to say that the most disadvantaged students attend schools with peers in similar situations and vice versa. High social segregation is also evident when looking at the distribution of students from different SES among different types of school providers. In Chile, for more than 20 years, schools have operated under a voucher system where parents can freely choose a school of their preference. Under this system, there have been three main types of providers of K-12 education: municipal schools (MUN), which provide public education and operate with public funding; private subsidized schools (PS), which provide private education and operate either with a combination of public and private funding (with tuition) or only with public funding (free of pay); and private paid schools (PP), which provide private education and operate with private funding (with tuition). Among these three types of providers, most underserved students are concentrated in MUN schools and wealthier students are concentrated in PP schools. Figure 2 shows how the distribution of students among the three types of schools’ providers varies according students’ family income. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 13 Source: Self elaboration with data from CASEN 2011. Figure 2. Distribution of Students enrolled in Private Paid, Private Subsidized and Municipalized Schools by Students’ Family Income In regards to academic segregation, Duarte et. al. (2013) found that, even though there have been positive changes in the distribution of learning achievement according to student SES from 1999 to 2011, the magnitude of these changes varies according to grade level and subject. The largest significant decrease in academic segregation is in fourth grade for the language test. There were no significant changes in eighth grade and a small but significant increase in academic segregation in tenth grade in both language and math. These findings suggest that a skimming process occurs as we move to higher grades. One of the consequences of the inequity problem at K-12 is that it will limit the opportunities for underserved students to continue to college. Since universities strongly rely on the PSU to select their students and students’ performance in this test is strongly correlated with students’ SES, universities have perpetuated the inequities that come from the K-12 education (Duarte et al., 2013; Hastings, Neilson, & Zimmerman, 2013; Koljatic, Silva, & Cofre, 2013). Consequently, there is a high disproportion of students who come from the top income quintile in university education. Seventy percent of the students coming from PP schools who take the PSU PREDICTING ENGINEERING STUDENT’S PERFORMANCE 14 are enrolled in one of the 33 universities that participate in the centralized admission process, compared with 32% for PS and 23% for MUN schools (Departamento de Evaluación, Medición, y Registro Educacional [DEMRE], 2014). This is important because the most prestigious and selective universities in Chile participate in this centralized admission process (Figure 3). * The data includes only the universities that adhere to the admission process administrated by DEMRE. These are: UCH, PUCCH, UDEC, PUCV, UTFSM, USACH, UACH, UCN, UV, UMCE, UTEM, UTA, UNAP, UANTOF, ULS, UPLA, UATA, UBB, UFRO, ULLAG, UMAG, UTAL, UCM, UCSC, UCT, UDP, UM, UFT, UAB, UAI, UANDES, UDD, UAH. Source: DEMRE (2014). Statistical Summary for the 2014 Admission Process. Figure 3. Students who Take the PSU and are Enrolled in a University by School’s Provider Given social and academic segregation at the school level, universities can play an important role in offering more opportunities to students from disadvantaged backgrounds. Furthermore, providing underserved students more opportunities would benefit both disadvantaged and advantaged students. There is evidence that a more diverse student body has a positive impact on students’ learning in addition to fostering personal development and active thinking for all students (Green, 2004; Gurin, Dey, Hurtado, & Gurin, 2002). Importantly, in the past years, some Chilean universities incorporated alternative selection methods. Most of these programs consider the student’s school ranking and SES as a central part of the selection process. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 15 In 2010, the PUC School of Engineering launched the T+I Program, which, in the following years, was scaled to the entire university. The program’s main goal is to attract talented students who experienced unequal opportunities during their primary and secondary education and were not able to access the PUC via the regular admission process. From 2011 to 2015, the T+I Program gave access to the school of engineering along with economic, academic and psychosocial support to more than 400 students. Statement of the Problem The alternative selection methods that several universities currently utilize consider, basically, the student’s SES and class ranking. There is broad literature that supports the relevance of high school grades and school ranking in predicting student success (Baron & Norman, 1992; French, Immekus, & Oakes, 2005; Harackiewicz, Barron, Tauer, & Elliot, 2002; Martin, Montgomery, & Saphian, 2006). Still, there is a gap in the literature in regards to other factors that could predict student performance and serve for selection purposes, particularly for engineering students. This question has been particularly interesting for the school of engineering at the PUC, which has been including, in addition to student’s class ranking and SES background, supplementary questionnaires and tests in the admission process for students entering via T+I Program, but with mixed results regarding how well these measures predict students’ performance. There is evidence indicating that field dependence-independence (FDI), logical- mathematical intelligence (LMI) and task persistence (TP) are attributes that correlate with STEM abilities and perseverance, both crucial components to succeed at a demanding, oftentimes stressful, and STEM-based career as engineering (Bowlin, 1988; Cerda, Ortega Ruiz, Pérez Wilson, Flores, & Melipillán, 2011; French, 1948). Nevertheless, no study has been done PREDICTING ENGINEERING STUDENT’S PERFORMANCE 16 utilizing these attributes to predict outcomes for engineering students in Chile, and, therefore, this study aimed to provide additional information in this regard to Chilean universities interested in predicting engineering students’ outcomes from non-traditional measures. Purpose of the Study The purpose of the study was to explore the relationship among students’ cognitive style of FDI, students’ LMI and students’ TPwith first-year performance at the PUC School of Engineering. Particularly, the study aimed to answer the following research questions: 1. To what extent does Field Dependence-Independence (FDI) predict academic performance of engineering students during their first year of college? 2. To what extent does Logical-Mathematical Intelligence (LMI) predict academic performance of engineering students during their first year of college? 3. To what extent does Task Persistence (TP) predict academic performance of engineering students during their first year of college? There is evidence to believe that each of these three individual characteristics could successfully serve as predictors. First, FDI cognitive style has been found to relate to students’ overall achievement (Bowlin, 1988; Tinajero & Páramo, 1997; Witkin et al., 2004). FDI is conceptualized as students’ differences in the ways they perceive and structure a field. Students who are more field-dependent (FD) will have a global perception where particular items are not well distinguished if they are presented in the context of a broader field. In contrast, more field- independent (FI) students will have an articulated perception where particular items are easily identified when they are presented in the context of a broader field (Demick, 2014; Witkin, Oltman, Raskin, & Karp, 1971). PREDICTING ENGINEERING STUDENT’S PERFORMANCE 17 Second, LMI has been also associated with overall higher achievement and, particularly, higher mathematical achievement (Nunes, Bryant, Barros, & Sylva, 2012; Nunes et al., 2007). LMI is conceptualized as the student’s ability of reasoning whether deductively or inductively; it is a student’s ability to confront the unknown by starting from what is known in terms of mental operation and, then, go to the unknown whose elements are expected to be equivalents (Etchepare, Ortega, Pérez, & Flores, 2011; Gardner & ebrary, 1999, 2000). Finally, it is expected that a more task persistent student would give up less frequently and this would have a positive impact on students’ grades (Andersson & Bergman, 2011; Boe, May, & Boruch, 2002; French, 1948). TP is defined as the students’ ability to persist at a task, as opposed to giving up, even if the task is perceived as excessively challenging. This conceptualization of persistence is often confused with perseverance, which is seen in this study as a longer-term attribute in terms of perseverance within one’s studies until graduation as opposed to quitting half way or dropping out (French, 1948). To answer the research questions, students in the sample were tested in FDI, LMI, and TP in January 2015, just after they were admitted to the school of engineering and before they started the school year in March. The instruments used to assess FDI, LMI and TP were, respectively, the Group Embedded Figure Test - Computerized Version (Witkin et al., 1971), the TILS, which in Spanish stands for Test de Inteligencia Lógica Superior (Etchepare et al., 2011), and the Ingenuity Test (French, 1948). The students’ results in these tests were compared with their performance in the different courses during the first and second semesters of 2015. The sample was made up of almost all students admitted via the T+I Program and a sample of students admitted via regular admission. The data were analyzed utilizing different statistical techniques, first without controls and, then, controlling for PSU. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 18 Significance of the Study The most important universities in Chile have recognized the deficiencies of an admission system that closes the doors to many talented students who come from more disadvantaged contexts. This problem is especially visible in a highly selective career such as engineering. This study aimed to provide additional evidence regarding the relevance of TP, FDI and LMI in predicting future performance of engineering students, particularly in the context of the Chilean population. Additionally, because students applying to engineering have a specific set of skills as compared to the general student population, this study may provide especially valuable information for all schools offering STEM-based programs. Furthermore, there is no study that focused on testing engineering students on these non-academic constructs in the context of the Chilean population. Finally, and considering external validity constraints, the findings of this study could serve to enlighten policy discussion and program design at a higher level. Limitations and Delimitations It is possible to identify several limitations that may have influenced the results of the following study. In regards to internal validity, the most important limitations are the following: 1. Sample size. The sample size of students is small and, therefore, even if there is a relationship between TP, FDI, and/or LMI and students’ future performance, a small sample size may make its detection difficult. 2. Instrumentation. There are three possible instrumentation problems. The first problem regards the measure of course grades and GPA for the first and second semester. In this case, the problem of instrumentation arises because of the multiple course sections that the school of engineering has for freshman students. This school enrolls PREDICTING ENGINEERING STUDENT’S PERFORMANCE 19 over 700 students each year and allocates students in more than 10 different course sections for each of the courses students take the first year (i.e., Calculus I., Linear Algebra, etc.). Professors in those sections design their own quizzes and midterm exams while the final exam is common to all sections. The second problem relates to the measure of FDI and LMI. The instruments chosen to measure these attributes may have ceiling effects since both were normalized for the general population, and engineering students are usually t the high end of the spectrum in regards to the ability measured in those tests. The third problem is related to the test chosen to measure TP. This test has not been validated and its use is exploratory. 3. Self-Selection bias. This could be present for the sample of students who entered via regular admission. While the group of students entering via the T+I Program took the tests in the context of a mandatory summer camp developed for them, such an opportunity did not exist to apply the test to students entering via regular admission. Multiple logistical implementation constraints only allowed invitation of students who registered on the first day of the inscription period. From that group of about 150 students, only 58 students attended the day of the tests. These 58 students may differ from the average population, as they may have a higher sense of responsibility or be more motivated and, therefore, affect the results of the study. 4. Omitted variables. There could be omitted variable bias since there may exist other variables, besides the one included in the models, that influence the relationship between FDI, LMI and/or TP and students’ future performance. In regards to external validity, this study is only generalizable to engineering students at the PUC. However, we know that the characteristics of engineering students at other universities PREDICTING ENGINEERING STUDENT’S PERFORMANCE 20 should be similar. Also, a broader interpretation or the results of this study may apply to students at other STEM careers. However, further exploration needs to be done in this regards. Organization of the Study The current chapter provided an overall introduction of the problem under study, its relevance for the Chilean society and the purpose of this study. Chapter Two presents the literature review, which discusses the structure of the Chilean higher education system in general and, in particular, the admission process at the PUC School of Engineering and the role of the T+I Program in this admission system. That chapter also reviews previous research in regards to TP, FDI and LMI. Chapter Three presents the methodology utilized to answer to the research questions and includes a description of the population sample as well as the instruments, the data collection method and the analysis technique. Chapter Four presents the results of the study, and, finally, Chapter Five discusses the most important findings of the study and the conclusion. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 21 CHAPTER TWO: LITERATURE REVIEW The following literature review is organized as follows: presented first is the structure of the Chilean higher education system in terms of the institutions that provide education, the type of degrees available at each of those institutions and the selection process. It is followed by a description of the admission process at the PUC School of Engineering and the role of the T+I Program within it. Finally, it presents a review of previous research in regards to TP, FDI and LMI and their relationship with student performance. Chilean Higher Education’s Institutions, Programs and Admission System Four types of institutions are the main providers of higher education in Chile (Gonzalez, 2005). Universities are licensed to award technical and professional degrees as well as academic degrees such as master’s and doctorates. Professional institutes, which have a more restricted license, are allowed to award technical and professional degrees. Technical training centers, are only licensed to award technical degrees. Educational institutions from the armed forces and order security are governed by their own statutes. There are four most common degrees available for students. Technical degrees are 2- years programs that require prior secondary school completion. Professional degrees consist of 5- or 6-year programs, depending on the career curriculum and also prior secondary school completion. Master’s degrees are 2-years programs that require prior completion of a professional degree, or, in some programs/universities, students can earn a masters’ degree concurrent with a professional degree after completing of the first 4 years of study and passing the bachelor’s exam. Doctoral degrees are earned through 4-year program that require previous completion of a professional degree, or, in some programs/universities, students can earn their PREDICTING ENGINEERING STUDENT’S PERFORMANCE 22 doctorate concurrent with their professional degree after completing of the first 4 years of study and passing the bachelor’s exam. Currently, there are over 200 institutions of higher education in Chile and, among them, are about 60 universities. Thirty-three universities, which include the top-ranked universities in Chile, participate in a centralized admission system administrated by the University of Chile. In this centralized system, students apply simultaneously to their preferred program/university, being able to rank up to eight preferences. For example, they can apply to engineering at PUC as first preference, to engineering at the University of Chile as second preference, to architecture at PUC as third preference and so forth. Finally, the student is accepted into only one of his/her preferences based on his/her final score. The final score on a student’s application is calculated differently by each program/university, which has the autonomy to decide the weights of each of the components in the final score, such as PSU scores (mathematics, science, language, history, etc.), school grades and school ranking (Hastings et al., 2013; Proceso de Admisión, n.d.). Pontificia Universidad Católica de Chile’s School of Engineering According to the QS Ranking 2015, the PUC School of Engineering is ranked 169 in the world, making it the best school in the country in engineering and technology and the third in South America. This highly selective program/institution selected about 750 students in 2015. The admission system to the school of engineering at the PUC considers applications from three different channels: regular admission, special admission, and admission via T+I Program. To apply via regular admission system, students must have completed secondary education and achieved a PSU score high enough to allow them admission. Regular admission is the most common way of entering engineering, and about 83% of the students entered this way in 2015. The second way is via special admission, which is designed for three types of students: students PREDICTING ENGINEERING STUDENT’S PERFORMANCE 23 with disabilities (auditory, motor or visual) who require special conditions to take the PSU; students who completed higher studies than secondary education (for example, students who want to move from engineering at another university to PUC); and students who excelled at the national level in the scientific, athletic, or artistic fields. In 2015, about 8% of students entered via the special admission channel. The last way of entering the PUC School of Engineering is via the T+I Program, which allows entrance to students who come from disadvantaged socioeconomic backgrounds, meaning from the lowest four income quintiles and who studied in a MUN or PS school. In 2015, 9% of students entered via the T+I Program (Admisión UC, n.d.). The Talent and Inclusion Program at PUC In 2010, the PUC School of Engineering launched the T+I Program, which, in the following years, was scaled to the entire university. The program’s main goal is to attract talented students who, because of a deficient K-12 education or socioeconomic problems, cannot access PUC via the regular admission process. To apply to the PUC through this program, students must fulfill several prerequisites. Among them, they need to rank in the top 10% of their class or have participated in PentaUC (a gifted program) for at least 11th and 12th grade, come from a MUN or PS school (MUN and PS schools are similar to the U.S. traditional public schools and charter schools, respectively), obtain the PSU minimum score required, and belong to one of the bottom four income quintiles. From 2011 to the time of this study, the T+I Program has given access to the school of engineering along with economic, academic and psychosocial support to more than 400 students. The role of the T+I Program in providing access to a body of more disadvantaged students is relevant at a national level. As previously presented, there is a large disproportion of students who come from the top income quintile in university education. Seventy percent of the PREDICTING ENGINEERING STUDENT’S PERFORMANCE 24 students coming from private schools who take the PSU are enrolled in one of the 33 universities that participate in the centralized admission process, as compared to 32% of students coming from PS schools and 23% of students coming from MUN schools. In this sense, the T+I Program has expanded opportunities to students who have the ability but have not had the same opportunities as their wealthier peers had. The PUC School of Engineering Curriculum Historically, engineering has been a 6-year program, with the first 3 to 4 years consisting of common core courses and the last 2 to 3 years made up of concentration courses. The current curriculum was implemented in 2013 as a way to offer students a more dynamic and flexible program. The curriculum was designed to be competitive at the international level. Therefore, the program enhances mobility to both local and international postgraduate programs and also allows students to accelerate their entrance into the job market and entrepreneurship. The curriculum promotes the applied sciences, technology, research and innovation among students. The PUC School of Engineering’s curriculum of is composed of two stages: the first stage leads to a Bachelor of Science in Engineering (4 years), and the second stage leads to the title of civil engineer, other professional title, a higher academic degree or early employment and entrepreneurship. First stage: Bachelor of Science in Engineering The degree is structured in a T Model, with a breadth component (interaction with several areas) and a depth component (area of concentration). The breadth component consists of the base courses in the fields of chemistry, biology, mathematics and physics; it also includes applied training in design, research, innovation and entrepreneurship. The depth component consists of courses relating to the major, and students have about 21 majors to choose from. In PREDICTING ENGINEERING STUDENT’S PERFORMANCE 25 addition to the previous courses, students need to take courses conducive to a minor; the minor can be part of the depth or the breadth components depending on whether the courses were respectively chosen in the same area as the major or in a different one. After four years, students can earn a Bachelor of Science in Engineering. Second Stage: Graduate Degree, Early Employment or Entrepreneurship After completing the bachelor’s degree, students have five options (Figure 4). They can exit to the job market or entrepreneurship. They may choose to continue to earn a professional degree in engineering at PUC, and students who choose this option are automatically accepted. Alternately, they may continue to earn another professional degree at PUC. Currently, there are agreements to continue studies leading to careers as surgeon, architect and designer. Students who choose this option need to apply, and vacancies are limited. Students may also continue to earn a graduate degree (master’s and/or doctorate) at PUC. Students who choose this option can simultaneously obtain the master’s or doctoral degree and the professional title of civil engineer at PUC. Lastly, they may continue to earn a graduate degree (master’s and/or doctorate) at a different university. Students who choose this option need to independently apply to other national or international universities. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 26 1 st Semester 2 nd Semester Subsequent 6 semesters Bachelor of Science in Engineering Exit to the Job Market or Entrepreneurship Calculus I* Calculus II 3 more years of courses conducting to the degree of Bachelor of Science in Engineering Civil Engineer from the PUC Leveling in Pre-Calculus General Chemistry II Statics and Dynamics Another Professional Degree from the PUC Leveling in General Chemistry Linear Algebra* Statics and Dynamics Lab Professional degree + Master degree at the PUC Leveling in Physics Challenges of Engineering Introduction to Economics Professional degree + Doctorate degree at the PUC Development of Communica- tional Skills OR Ethics Development of Communica- tional Skills OR Ethics Master’s or Doctorate degree at another University * Prerequisite courses for Calculus II and Statics and Dynamics Source: self elaboration based on information provided by the School of Engineering website: http://www.ing.uc.cl/alumnos/plan-de-estudios/descripcion/ Figure 4. Curriculum for the Bachelor’s Degree in Science of Engineering Field Dependence-Independence, Logical-Mathematical Intelligence, Task Persistence The following section presents a review of the literature in regards to FDI, LMI, and TP. It describes how these three concepts are conceptualized for this study and presented previous research that supports the premise that FDI, LMI and TP are good predictors of student performance. Field Dependence-Independence Witkin defined FDI as a cognitive style in the 1960’s. Witkin (1967) defined cognitive style as the “characteristic self-consistent modes of functioning found pervasively throughout an individual's cognitive, that is, perceptual and intellectual, activities” (p. 234). Cognitive style can PREDICTING ENGINEERING STUDENT’S PERFORMANCE 27 also be understood as “a person’s characteristic and usually preferred ways of processing information” (Sternberg & Grigorenko, 1997, p. 700). In particular, Witkin was interested in the cognitive style of FDI which describes differences in the ways of perceiving and structuring a field: Field dependent (FD) performance is characterized by global perception where items are not well differentiated from the background when the field is structured and where there is a lack of the imposition of organization on the field when it is unstructured. In contrast, field independent (FI) performance is characterized by articulated perception where items are readily disembedded from the surrounding field and where structure is imposed on an unstructured field. (Demick, 2014, p. 13). FDI has broadly been studied in regards to educational outcomes and has been found to correlate to students’ achievement. Witkin et al. (1977) longitudinally followed 1,548 students from college entry until they entered graduate/ professional school. They tested students on the Group Embedded Figures Test (GEFT) at college entry and evaluated if their cognitive style was related to their affinity for different majors as well as their academic performance. They found that students who were more FI had more affinity for impersonal domains such as sciences and that who were more FD had more affinity for interpersonal domains such as elementary education. Interestingly they found also that students who do not choose a major that fits their cognitive style tend to shift to more compatible domains. They also found some evidence that students tend to do better in domains that are compatible with their cognitive style. Finally, they found that the GEFT correlates with students’ performance in mathematic and sciences but that this correlation is not significant when controlling for SAT scores. Another study by Bowlin (1988) tested 100 senior college students in FDI and found a low, but significant, correlation PREDICTING ENGINEERING STUDENT’S PERFORMANCE 28 between FDI and students’ performance in math, social studies, and science for those who tested FI as well as a low, but significant, correlation between FDI and IQ scores. Finally, Tinajero and Páramo (1997) examined the relationship between academic achievement and FDI in 408 13- to 16-year-old students. They found that FI students performed better than did FD ones in most of the subjects measured, including mathematics and overall achievement. Another interesting outcome of this study is that the authors tested the students using the Rod and Frame Test and the Embedded Figures Test, and the significant results were mostly found through the Embedded Figures Test. Logical-Mathematical Intelligence According to Gardner and ebrary (1999, 2000) there are seven different kinds of human intelligence: linguistic intelligence, logical-mathematical intelligence, musical intelligence, bodily-kinesthetic intelligence, spatial intelligence, interpersonal intelligence and intrapersonal intelligence. Among them, “logical-mathematical intelligence involves the capacity to analyze problems logically, carry out mathematical operations, and investigate issues scientifically” (Gardner & ebrary, 1999, p. 42). LMI can be also conceptualized as the student’s ability of reasoning, whether deductively or inductively, to confront the unknown by starting from what is known in terms of mental operation and then go to the unknown whose elements are expected to be equivalents (Etchepare et al., 2011; Gardner & ebrary, 1999, 2000). According to Cerda et al. (2011), LMI is critical in the development of reasoning, deduction, abstract thinking and resolution of problems. In a study of the Chilean population, Cerda et al. (2011) tested 4446 students from both elementary and secondary education in logic intelligence and found a positive and significant relationship between LMI and overall academic achievement, especially in mathematics. They also found that LMI differs significantly among PREDICTING ENGINEERING STUDENT’S PERFORMANCE 29 students depending on their age, gender and school provider (as a measure of socioeconomic background). In another study, Korkmaz (2012) studied 45 first-year students at the Department of Computer and Instruction Technologies in an education school as to how their critical thinking and LMI predicted their ability to design algorithms. He found that both students’ logical- mathematical intelligence and critical thinking abilities positively influence their algorithm design skills and that jointly explained 59.7% of the variance of algorithm design skills. Finally, Callaman (2014) assessed 217 fourth-year high school students on intellective and non- intellective factors, including LMI, family income and parent’s education, among others, to evaluate the relationship between those factors and students’ performance in advanced algebra. He concluded that LMI positively correlates with student achievement in understanding, applying, analyzing and evaluating advanced algebra. Algebra and mathematics in general as well as computer programming are important subjects within the engineering curriculum, and, therefore, these studies provide important evidence in regards to how LMI could positively predict engineering students’ performance. Task Persistence TP can be defined as “the ability to persist and to sustain attention at a task, even in the presence of internal and external distractions” (Andersson & Bergman, 2011, p. 950). The present study complements the previous definition with the conceptualization from French (1948), who defined TP as the individual’s ability to persist at a task, as opposed to giving up, even if the task is perceived as excessively challenging. Several studies identified TP as a critical component of performance. Boe et al. (2002) used the Third International Mathematics and Science Study (TIMSS) to analyze the relevance of TP on students’ performance. Student task persistence (STP) was measured as an index of PREDICTING ENGINEERING STUDENT’S PERFORMANCE 30 student engagement in providing answers to TIMSS questions on the background questionnaire. Using multilevel analyses, they studied the relationships between STP and achievement at the student, classroom, and national levels. They concluded that STP is one of the strongest predictors of national differences in mathematics and science achievement, explaining more than 50% of the variance at the national level. In another study, Anderson and Bergman (2011) used data from the Swedish longitudinal research program to analyze the influence of TP in educational and occupational attainment in middle adulthood. They found that TP measured at the age of 13 significantly predicts students’ performance at the age of 16 and educational attainment in middle adulthood. Additionally, French (1948) performed a study on freshman engineering students entering the Cooper Union Engineering School. To test TP, he used number series problems where some problems had no solution. Students were informed about the existence of problems without a solution, but they were unaware that the goal of the test was to assess them on TP. Therefore, the test evaluates the capacity of the students to persist on each problem and not mark them as “without solution.” French (1948) found that the results of the test positively correlated with students’ grades. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 31 CHAPTER THREE: METHODOLOGY The present study utilized inferential statistics to evaluate the extent to which TP, FDI and LMI can predict engineering students’ performance during their first year of college. Population and Sample The population of the study can be defined from two different approaches. A more conservative approach would define the population only as the students enrolled at the PUC School of Engineering because the sample was drawn from that school only and, therefore, results are only generalizable to those students. A less conservative approach could consider that students entering engineering share similar ability profiles and that the first-year curriculum at PUC is similar to first-year curriculum in other engineering programs. Therefore, the total population of students and the generalizability of the study encompass all students in engineering programs at the different universities in Chile. Considering the first approach, the population would be 757 students, who were all the students enrolled at PUC in 2015. Considering the second approach, the population would be 26,368 students, who are all students enrolled in Civil Industrial Engineering and Civil Engineering Common Plan programs in 2015 (National Council of Education, 2015). The study defined a sample of students from the PUC School of Engineering, which initially considered all students entering via the T+I Program and a sample of students entering via regular admission. Finally, from the 757 students enrolled in engineering, the sample was composed of 65 students admitted via the T+I Program (from a total of 70) and 58 students admitted via regular admission (from a total of 628). To select the sample of students entering via regular admission, all students who registered on the first day of a three-day inscription PREDICTING ENGINEERING STUDENT’S PERFORMANCE 32 period were invited to participate. From that group of 150 students, 58 students attended to take the tests. Table 1 Students in the Sample from the PUC School of Engineering Sample Total T+I Program 65 70 Regular Admission 58 628 Special Admission 0 59 Total 123 757 Instrumentation The tests to assess FDI, LMI and TP were, respectively, the Group Embedded Figures Test - Computerized Version, also known as the GEFT-C (Witkin et al., 1971), the Test of Superior Logical Intelligence (TILS, Spanish for Test de Inteligencia Lógica Superior) (Etchepare et al., 2011) and the Numerical Ingenuity Test – TIN (in Spanish stands for Test de Ingenio Numérico) (French, 1948). Group Embedded Figures Test - Computerized Version (GEFT-C) The GEFT was developed by Witkin et al. (1971), and, for this study, the computerized version of the test (GEFT-C) was used. The GEFT was designed to identify students’ ability to separate detailed information from the context around it. In each problem, students have to isolate a common geometric shape in a larger design. Students have 20 minutes to answer 25 problems, which are divided into three sections. The scoring of the test is usually based on sections 2 and 3, making for a maximum score of 18. However, in this study, the three sections were scored, making for a maximum score is 25. The main reason to consider the three sections was that the test results where highly skewed, presenting a ceiling effect in an important number of cases. When considering the three sections, the number of students who scored the maximum score decreased from 30 (24.0%) to 21 (16.8%), increasing the validity of the measure. On this PREDICTING ENGINEERING STUDENT’S PERFORMANCE 33 test, the higher the score indicates a more FI student, and a lower the score indicates a more FD student, creating continuity between the two poles. The reliability estimate of the GEFT is 0.82 according to Witkin et al. (1971) for the paper version, and Kepner and Neimark (1984) report a correlation between the paper version and the computerized version of 0.78. Worthy of mention is that the test has been tested in different populations. Ngnoumen et al. (2014) report a mean for Ivy League undergrads of 14.5 and Zimmerman, Johnson, Hoover, Hilton, Heinemann and Buckmaster (2006) report a mean for students in engineering majors in a U.S. state university of 16.5. Figure 5. Example of the Type of Problems in the GEFT Test Test of Superior Logical Intelligence (TILS) The TILS was developed by Tejares from the San Pio X Institute, Salamanca, Spain and then normalized for the Chilean population by Riquelme, Segure and Yévenes (1991). The test was designed to assess students’ ability to reason whether deductively or inductively (Etchepare et al., 2011; Gardner & ebrary, 1999, 2000). Each problem shows a sequence composed of four figures wherein the student needs to determine, among five possible alternatives, which figure is PREDICTING ENGINEERING STUDENT’S PERFORMANCE 34 the one that continues the series. Students had 30 minutes to answer 50 problems. The TILS has shown to be highly reliable with a Cronbach Alpha of 0.95 according to Etchepare et al. (2011). Figure 6. Example of the Type of Problems in the TILS Test Numerical Ingenuity Test (TIN) The TIN was developed by French (1948) to test engineering students’ capacity to persisting at task in the sense of the student’s willingness to keep trying even when presented with a difficult task on which less persistent students would be tempted to give up. For the present study, the test was translated to Spanish under the name Test de Ingenio Numérico. The test consists of number series problems with different levels of difficulty, from simple to very difficult, with a few problems without solution. Students were told that some problems have no solution and that they need to write an X if they believe so, or leave it blank if they do not know the answer. In addition, students were asked their perception about the level of difficulty of each problem, which can be easy, moderate or difficult. Finally, to track the time and number of reversals, students also needed to register the time when they wrote or changed an answer. The time is presented in the front of the room with numbers from 1 to 60 that represents the time in minutes. Students have one hour to answer 30 problems. In his study, French (1948) did not present an internal validity estimator for the test. Therefore, Chapter Four presents the results Cronbach’s Alpha for different sub-dimensions of the TIN scale. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 35 1st 2nd 3rd 4th 5th 6th 7th 8th Difficulty Number Being Displayed 1 2 4 6 8 10 12 2 5 10 15 25 40 65 Figure 7. Example of the Type of Problems in the Numerical Ingenuity Test Data Collection Students in the sample were tested in TP, FDI and LMI in January 2015, just after they were admitted to the school of engineering and before they started the school year in March. The timing, logistics and methods for convocation of students were planned in collaboration with the school of engineering. Worth mentioning is that this study was considered a part of the School of Engineering Institutional Improvement Plan, and, therefore, the students were asked to participate by personnel from the PUC. Students from the T+I Program group were asked to take the tests in the context of a week of mandatory summer camp for students who entered via this program, and they were tested in two consecutive afternoons. The first day, they were administered the TIN, which was the most challenging test, and, the second day, they were administrated the GEFT-C and the TILS. Students entering via regular admission were tested in a different day. This group of students was invited to participate during the first day of a three-day inscription period. There was no methodological reason to invite only students registering the first day but merely logistic limitations. From about 150 students invited, 58 presented on the day of the test. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 36 Students were told that the tests were not part of their academic program and would not have any consequences on their future. They were also informed that the data would be analyzed in an aggregate manner and used for institutional improvement. Analysis The analysis of the data considers using statistical techniques to compare the results from the GEFT-C, TILS and all sub-dimensions of the TIN with students’ course grades, GPA from the first and second semester of 2015 and cumulative GPA after the first year of college. The analysis was carried out, first, with no control variables and then, when possible, controlling for PSU. Independent variables were measured in some cases with a unique sub-dimension and in other cases with more than one sub-dimension. For the cases of FDI and LMI, both were measured with a unique sub-dimension. The scoring for the GEFT-C (for measuring FDI) and the TILS (for measuring LMI) were straightforward and one final score was calculated. Nevertheless, the scoring for the TIN (for measuring TP) is more complex. French (1948) defined six variables that could be related to STP, finding mixed results for each of them: (1) the maximum amount of time the students spent in the test, (2) the amount of attempts per problem the students made, (3) the number of problems answered correctly, (4) the number of items which were not answered with an X, (5) the productivity during the second half hour of the test, and (6) the time for going through the test once. In this study, five of the six variables proposed by French (1948) were considered and a new variable meant to measure to what degree students attempted to guess the answers in the test was considered. Students’ guessing was measured by calculating the ratio between written wrong answers (blanks not included) and all wrong answers. This variable was 0 if all the written answers were correct (in other words, if all the PREDICTING ENGINEERING STUDENT’S PERFORMANCE 37 wrong answers are blanks) and was 1 if the student did not leave any blank answers (meaning that, if the student did not know an answer, he/she wrote an answer anyway). Finally, the six variables analyzed in regards to the TIN were number of items answered correctly, number of items answered with an X, highest time-number, number of items answered during the secnd half hour, student guessing and total number of reversals. The variables that were explored in the analysis are the following: Dependent variables: • Cumulative GPA: students’ cumulative GPA in the first year. • 1 st Semester GPA: students’ GPA in the first semester. • 2 st Semester GPA: students’ GPA in the second semester. • Calculus I: students’ final grade in Calculus I, a class taken in the first semester. • Chemistry II: students’ final grade in Chemistry II, a class taken in the first semester. • Linear Algebra: students’ final grade in Linear Algebra, a class taken in the first semester. • Challenges Engineering: students’ final grade in Challenges of Engineering, a class taken in the first semester. • Calculus II: students’ final grade in Calculus II, a class taken in the second semester. • Statics & Dynamics: students’ final grade in Statics and Dynamics, a class taken in the second semester. • Economics: students’ final grade in Economics, a class taken in the second semester. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 38 • Programming: students’ final grade in Programming, a class taken in the second semester. • Comm. Skills: students’ final grade in Communication Skills, an elective class taken in the first year. • Ethics: students’ final grade in Ethics, an elective class taken in the first year. Independent variables: • Corrects in GEFT-C: A measure of FDI. • Corrects in TILS: A measure of LMI. • Reversals in TIN: A possible measure of TP. • Corrects in TIN: A possible measure of TP. • Mark with X in TIN: A possible measure of TP. • Max time mark in TIN: A possible measure of TP. • Number of items answered during the 2nd half hour in TIN: A possible measure of TP. • Student Guessing in TIN: A possible measure of TP. Student guessing is calculated as the ratio between written wrong answers (including X’s) and wrong answers in the TIN test. • PSU total score: final weighted score in the PSU. This is a control variable. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 39 CHAPTER FOUR: RESULTS The following chapter presents the results of the study. The data were analyzed following the techniques described in Chapter Three, keeping the focus on answering the previously stated research questions: 1. To what extent does Field Dependence-Independence (FDI) predict academic performance of engineering students during their first year of college? 2. To what extent does Logical-Mathematical Intelligence (LMI) predict academic performance of engineering students during their first year of college? 3. To what extent does Task Persistence (TP) predict academic performance of engineering students during their first year of college? FDI was measured with the GEFT-C, LMI was measured with the TILS and TP was measured through different sub-dimensions of the TIN. Chapter four is organized as follows. First, the descriptive statistics for the dependent and independent variables are presented. Second, the results that account for the relationship between FDI and students’ performance are presented, with the aim of answering the first research question. Third, the results in regards to the relationship between LMI and students’ performance are presented with the aim of answer the second research question. Finally, the results that account for the relationship between TP and students’ performance are presented with the aim of answering the third research question. Descriptive Statistics Dependent Variables Descriptive data for the dependent variables, including sample size, range, mean, variability and measures for assessing the shape of the distribution, are presented in Table 2. In PREDICTING ENGINEERING STUDENT’S PERFORMANCE 40 regards to sample size, this varies depending on whether the course is from the first or second semester. Three courses from the second semester have a considerably lower sample size: Calculus II, Statics and Dynamics and Programming. There are reasons for this; in the case of Calculus II, it is necessary to pass Calculus I in order to register for Calculus II, and, in the case of Statics and Dynamics and Programming, it is possible that students who did not pass one or more classes from the first semester postponed registering in those courses. In regards to the range, all dependent variables have a theoretical range from 1 to 7 because the Chilean system measures performance on that scale, with 4 (3.95) the minimum grade for passing. As is seen in Table 2, the grade means for the different courses range between 3.57 and 5.41, with the mean for the first semester GPA at 4.50. For the second semester, the GPA is at 4.85, and, for the first year, the cumulative GPA is at 4.72. In regards to the distribution of the dependent variables, many of them had high levels of skewness. Chemistry II, Statics and Dynamics, Economics, Programming, Communication Skills, Ethics, and first and second semester GPA have skewness z statistics larger than 1.96, suggesting significant deviations from normality. Only Calculus I, Calculus II, Linear Algebra and cumulative GPA distribute normally. Figures 8 to 11 show the histograms for the dependent variables wherein these problems can be observed visually. Table 3 shows the correlations among dependent variables. Even though most of the dependent variables show significant positive correlations among them, a notably higher correlation exists among mathematics-oriented courses. The largest correlations are found among Calculus I, Chemistry II, Linear Algebra, Calculus II and Statics and Dynamics, with the highest correlation between Calculus I and Linear Algebra, r=0.827, p<0.001. In contrast, Communication Skills and Ethics have the smallest correlations with all the other courses, showing medium to small correlations, but, still, many of them are significant. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 41 Table 2 Descriptive Data for Dependent Variables N Min Max Mean Std. Dev. Var. Skewness Kurtosis Stat. Stat. Stat. Stat. Std. Error Stat. Stat. Stat. Std. Error z Stat. Stat. Std. Error z Stat. Classes First Semester Calculus I 123 1.8 6.2 4.342 0.091 1.005 1.011 -0.341 0.218 1.564 -0.514 0.433 1.187 Chemistry II 123 1.0 6.5 4.346 0.114 1.265 1.6 -0.587 0.218 2.693 -0.403 0.433 0.931 Linear Algebra 123 1.0 6 3.570 0.102 1.133 1.285 -0.132 0.218 0.606 -0.838 0.433 1.935 Challenges Engineering 123 2.2 7 5.042 0.050 0.549 0.301 -0.389 0.218 1.784 7.025 0.433 16.224 Classes Second Semester Calculus II 78 2.5 6.6 4.840 0.086 0.761 0.579 -0.439 0.272 1.614 0.477 0.538 0.887 Statics & Dynamics 67 1.1 6 4.164 0.148 1.214 1.474 -0.805 0.293 2.747 -0.333 0.578 0.576 Economics 102 2.5 7 5.273 0.072 0.726 0.527 -0.58 0.239 2.427 1.094 0.474 2.308 Programming 64 1.1 6.5 5.023 0.136 1.084 1.175 -1.149 0.299 3.843 1.961 0.59 3.324 Elective Classes First Year Comm. Skills 105 1.0 6.8 5.407 0.091 0.929 0.864 -1.871 0.236 7.928 5.755 0.467 12.323 Ethics 110 2.5 6.7 5.391 0.068 0.716 0.513 -0.796 0.23 3.461 1.593 0.457 3.486 Grade Point Average GPA 1 st Sem 123 1.5 6.1 4.501 0.070 0.778 0.605 -0.517 0.218 2.372 0.658 0.433 1.520 GPA 2 nd Sem 115 2.9 6.1 4.854 0.066 0.705 0.496 -0.444 0.226 1.965 -0.431 0.447 0.964 Cum. GPA 115 3.0 6.1 4.722 0.056 0.605 0.366 -0.144 0.226 0.637 -0.157 0.447 0.351 Calculus I Chemistry II Linear Algebra Challenges of Engineering Figure 8. Histograms for Dependent Variables: Mandatory Classes 1 st Semester Frequency Frequency PREDICTING ENGINEERING STUDENT’S PERFORMANCE 42 Calculus II Statics and Dynamics Economics Programming Figure 9. Histograms for Dependent Variables: Mandatory Classes 2 nd Semester Communication Skills Ethics Figure 10. Histograms for Dependent Variables: Elective Classes 1 st Year Frequency Frequency Frequency PREDICTING ENGINEERING STUDENT’S PERFORMANCE 43 GPA 1 st Semester GPA 2 nd Semester Cumulative GPA 1 st year Figure 11. Histograms for Dependent Variables: Grade Point Average and Cumulative Grade Point Average Frequency PREDICTING ENGINEERING STUDENT’S PERFORMANCE 44 Table 3 Spearman's rho Correlations Coefficients for DV’s Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 1. Calculus I r 1 Sig. . N 123 2. Chemistry II r 0.767** 1 Sig. 0.000 . N 123 123 3. Linear Algebra r 0.827** 0.819** 1 Sig. 0.000 0.000 . N 123 123 123 4. Challenges Engineering r 0.282** 0.441** 0.391** 1 Sig. 0.002 0.000 0.000 . N 123 123 123 123 5. Communication Skills r 0.319** 0.177 0.305** 0.204* 1 Sig. 0.001 0.071 0.002 0.037 . N 105 105 105 105 105 6. Ethics r 0.297** 0.307** 0.285** 0.376** 0.155 1 Sig. 0.002 0.001 0.003 0.000 0.141 . N 110 110 110 110 92 110 7. Calculus II r 0.662** 0.600** 0.734** 0.318** 0.138 0.356** 0.003 69 1 Sig. 0.000 0.000 0.000 0.005 0.258 . N 78 78 78 78 69 78 8. Statics & Dynamics r 0.656** 0.678** 0.736** 0.325** -0.013 0.194 0.676** 1 Sig. 0.000 0.000 0.000 0.007 0.925 0.135 0.000 . N 67 67 67 67 58 61 67 67 9. Economics r 0.496** 0.468** 0.536** 0.476** 0.344** 0.320** 0.575** 0.546** 1 Sig. 0.000 0.000 0.000 0.000 0.001 0.002 0.000 0.000 . N 102 102 102 102 90 93 70 60 102 10. Programming r 0.484** 0.526** 0.601** 0.322** 0.445** 0.356** 0.536** 0.506** 0.497** 1 Sig. 0.000 0.000 0.000 0.010 0.001 0.006 0.000 0.000 0.000 . N 64 64 64 64 53 59 52 49 55 64 11. 1st Semester GPA r 0.881** 0.906** 0.922** 0.533** 0.369** 0.445** 0.720** 0.695** 0.589** 0.612** 1 Sig. 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 . N 123 123 123 123 105 110 78 67 102 64 123 12. 2nd Semester GPA r 0.412** 0.390** 0.550** 0.298** 0.460** 0.087 0.659** 0.676** 0.656** 0.684** 0.499** 1 Sig. 0.000 0.000 0.000 0.001 0.000 0.378 0.000 0.000 0.000 0.000 0.000 . N 115 115 115 115 102 105 78 67 102 64 115 115 13. Cumulative GPA r 0.740** 0.727** 0.846** 0.451** 0.420** 0.271** 0.791** 0.757** 0.741** 0.707** 0.854** 0.862** 1 Sig. 0.000 0.000 0.000 0.000 0.000 0.005 0.000 0.000 0.000 0.000 0.000 0.000 . N 115 115 115 115 102 105 78 67 102 64 115 115 115 **. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed). Independent Variables Descriptive data for independent variables are shown in Table 4, including sample size, range, mean, variability and measures for assessing the shape of the distribution. Sample size for all independent variables is similar, from 122 to 125 students. Independent variables’ range varies depending on the test: the theoretical range for the GEFT-C is 0-25. For TILS, it is 0-50, and, for TIN (Corrects in TIN), it is 0-30. For all of the previous tests, the minimum unit of PREDICTING ENGINEERING STUDENT’S PERFORMANCE 45 measurement is the integer. The theoretical range for the PSU is 150.00-850.00, and the minimum unit of measurement the hundredth. The PSU minimum score showed in Table 4 represents the minimum score obtained by a student entering via the T+I Program. However, it is worth mentioning that the cut score for engineering students entering via regular admission was 735.10. In regards to the distribution of independent variables, several of them do not normally distribute. All of the following show highly skewed distributions: number of corrects in GEFT- C, number of corrects in TILS, number of reversals in TIN, number of items marked with an X in TIN and maximum time marked in TIN. Only four out of nine showed normal distributions: number of corrects in TIN, number of items answered during the 2nd half hour in TIN, student guessing in TIN and PSU total score. Figures 1 to 15 show the histograms for the independent variables wherein these problems can be observed visually. Table 4 Descriptive Data for Independent Variables N Min. Max. Mean Std. Dev. Var. Skewness Kurtosis Stat. Stat. Stat. Stat. Std. Error Stat. Stat. Stat. Std. Error z Stat. Stat. Std. Error z Stat. Field Dependence-Independence (Group Embedded Figure Test - GEFT-C) Corrects in GEFT-C 123 9 25 21.54 0.312 3.455 11.9 -1.464 0.218 6.716 2.107 0.433 4.866 Logical-Mathematical Intelligence (Superior Logical Intelligence Test - TILS) Corrects in TILS 122 24 48 38.48 0.483 5.337 28.5 -0.582 0.219 2.658 -0.263 0.435 0.605 Persistence (Numerical Ingenuity Test - TIN) Reversals in TIN 125 0 5 0.27 0.070 0.787 0.6 3.919 0.217 18.060 17.382 0.430 40.423 Corrects in TIN 125 7 25 16.01 0.288 3.224 10.4 -0.154 0.217 0.710 0.476 0.430 1.107 Mark with X in TIN 125 0 14 2.74 0.247 2.759 7.6 1.761 0.217 8.115 3.654 0.430 8.498 Max time mark in TIN 124 24 60 54.70 0.518 5.766 33.3 -2.276 0.217 10.488 7.352 0.431 17.058 Number of items answered during the 2nd half hour in TIN 124 0 17 7.19 0.286 3.187 10.2 0.234 0.217 1.078 -0.054 0.431 0.125 Student Guessing in TIN 125 0 1 0.46 0.023 0.259 0.1 0.416 0.217 1.917 -0.539 0.430 1.253 PSU (University Selection Test) PSU total score 125 688.10 828.10 751.00 3.063 34.245 1172.8 0.042 0.217 0.194 -0.795 0.430 1.849 Corrects in GEFT-C PREDICTING ENGINEERING STUDENT’S PERFORMANCE 46 Figure 12. Histograms for Field Dependence-Independence (GEFT-C) Corrects in TILS Figure 13. Histogram for Logical-Mathematical Intelligence (TILS) Frequency Frequency PREDICTING ENGINEERING STUDENT’S PERFORMANCE 47 Reversals in TIN Corrects in TIN Marx with X in TIN Maximum Time in TIN Number of items answered during the 2nd half hour in TIN Student Guessing in TIN Figure 14. Histograms for Task Persistence (TIN) PSU Score Figure 15. Histogram for PSU Score Correlations among independent variables are shown in Table 5. The correlations among the final scores on the different tests, including the PSU, reveal mostly medium-size positive correlations, 0.29<r<0.39, p≤0.01; only the final scores on the GEFT-C and TILS have a large correlation, r=0.50, p<0.01. There are also large correlations between three sub-dimensions of the TIN test: “student guessing” had a correlation with “mark with a X” and “number of items Frequency Frequency Frequency PREDICTING ENGINEERING STUDENT’S PERFORMANCE 48 answered during the 2nd half hour” of 0.72 and 0.53 respectively (p<0.01); in addition, “number of items answered during the 2nd half hour” had a correlation with “max time marked” of 0.57 (p<0.01). Independent variables with large correlations may indicate multicollinearity problems, which will be considered in future analysis. Table 5 Spearman's rho Correlations Coefficients for IV’s Measure 1 2 3 4 5 6 7 8 9 1. Corrects in GEFT- C r 1 Sig. . N 123 2. Corrects in TILS r 0.501** 1 Sig. 0.000 . N 122 122 3. Reversals in TIN r -0.155 -0.074 1 Sig. 0.087 0.417 . N 123 122 125 4. Corrects in TIN r 0.301** 0.290** -0.041 1 Sig. 0.001 0.001 0.646 . N 123 122 125 125 5. Mark with X in TIN r -0.172 -0.171 0.094 -0.072 1 Sig. 0.057 0.060 0.300 0.428 . N 123 122 125 125 125 6. Max time mark in TIN r -0.082 -0.131 -0.033 0.062 0.150 1 Sig. 0.368 0.153 0.715 0.496 0.096 . N 122 121 124 124 124 124 7. Number of items answered during the 2nd half hour in TIN r 0.029 -0.185* -0.006 0.060 0.401** 0.571** 1 Sig. 0.748 0.042 0.943 0.509 0.000 0.000 . N 122 121 124 124 124 124 124 8. Student Guessing in TIN r 0.037 -0.030 0.060 0.113 0.723** 0.227* 0.532** 1 Sig. 0.684 0.745 0.503 0.211 0.000 0.011 0.000 . N 123 122 125 125 125 124 124 125 9. PSU total score r 0.341** 0.387** -0.026 0.391** -0.326** -0.176 -0.250** -0.240** 1 Sig. 0.000 0.000 0.774 0.000 0.000 0.051 0.005 0.007 . N 123 122 125 125 125 124 124 125 125 ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). Internal Consistency of the Tests Group Embedded Figure Test (Computerized version) – GEFT-C. According to Witkin et al. (1971) the reliability estimate of the GEFT is 0.82 for the paper version, and Kepner and Neimark (1984) report a correlation between the paper version and the computerized version of 0.78. For this study the GEFT-C showed to be highly reliable, with a calculated Cronbach’s PREDICTING ENGINEERING STUDENT’S PERFORMANCE 49 Alpha of 0.82, considering 25 items that correspond to the total of items, including sections 1, 2 and 3, and a sample of 123 students. Superior Logical Intelligence Test – TILS. Etchepare et al. (2011) reported a Cronbach’s Alpha of 0.95 for the TILS. For this study, considering the 50 items in the TILS and a sample of 123 students, the calculated Cronbach’s Alpha was 0.86, which still indicates a high level of internal consistency for the TILS scale. Numerical Ingenuity Test – TIN. Six sub-dimensions where measure in the TIN: number of items answered correctly, number of items answered with an X, highest time-number, number of items answered during the 2nd half hour, student guessing and total number of reversals. It was possible to calculate Cronbach’s Alpha for three sub-dimensions that have item level data, these are: number of items answered correctly, number of items answered with an X, and number of items answered during the 2nd half hour. Table 6 presents the Cronbach’s Alpha coefficients for these sub-dimensions, and, as is seen, the most reliable sub-dimension is the number of items answered with an X with a Cronbach’s Alpha of 0.75, which, though not very high, it does meet the general requisite to be higher than 0.7 (Pallant, 2013). In contrast, the other two sub-dimensions have Cronbach’s Alpha lower than 0.7; therefore, these sub-dimensions are probably measuring more than one underlying concept. Nonetheless, these sub-dimensions were still considered in the analysis, and results must be interpreted prudently. Finally, it seems pertinent to clarify that the reason to consider 28 items (instead of 30) to calculate Cronbach’s Alpha for the sub-dimension number of items answered with an X was that two of the questions had, indeed, no solution, and, therefore, marking X in these questions did not measure “lack of task persistence,” but the correct answer. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 50 Table 6 Cronbach’s Alpha for Sub-Dimensions in TIN N Number of Items Cronbach’s Alpha Number of items answered correctly 125 30 0.599 Number of items answered with an X 125 28 0.746 Number of items answered during the 2nd half hour 125 30 0.508 Field Dependence-Independence and Student Performance FDI results, measured through the GEFT-C, are highly negatively skewed, experiencing a ceiling effect in 21 of the 123 observations (16.8%). Given the lack of normality, the Spearman's rho correlation coefficient was used (Table 7). Positive small and significant correlations between FDI and Calculus I r(119)=0.180, p=0.048; Linear Algebra r(119)=0.260, p=0.004; 1st Semester GPA r(119)=0.217, p=0.017; and Cumulative GPA r(112)=0.208, p=0.027 are observed. Field Dependence-Independence as a Predictor of Students Performance To further analyze whether FDI could be a good predictor of performance, an independent-samples t-test was conducted to compare grades between students who were above and below the average in the FDI spectrum for all the courses during the first and second semester as well as 1st semester GPA, 2nd semester GPA and cumulative GPA. Table 8 shows the group statistics for the independent-samples, and Table 9 shows a summary of the t-test results. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 51 Table 7 Spearman's rho Correlations Coefficients for FDI (measured through the GEFT-C) and DV’s Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1. Calculus I r 1 Sig. . N 123 2. Chemistry II r 0.767** 1 Sig. 0.000 . N 123 123 3. Linear Algebra r 0.827** 0.819** 1 Sig. 0.000 0.000 . N 123 123 123 4. Challenges Engineering r 0.282** 0.441** 0.391** 1 Sig. 0.002 0.000 0.000 . N 123 123 123 123 5. Communication Skills r 0.319** 0.177 0.305** 0.204* 1 Sig. 0.001 0.071 0.002 0.037 . N 105 105 105 105 105 6. Ethics r 0.297** 0.307** 0.285** 0.376** 0.155 1 Sig. 0.002 0.001 0.003 0.000 0.141 . N 110 110 110 110 92 110 7. Calculus II r 0.662** 0.600** 0.734** 0.318** 0.138 0.356** 0.003 69 1 Sig. 0.000 0.000 0.000 0.005 0.258 . N 78 78 78 78 69 78 8. Statics & Dynamics r 0.656** 0.678** 0.736** 0.325** -0.013 0.194 0.676** 1 Sig. 0.000 0.000 0.000 0.007 0.925 0.135 0.000 . N 67 67 67 67 58 61 67 67 9. Economics r 0.496** 0.468** 0.536** 0.476** 0.344** 0.320** 0.575** 0.546** 1 Sig. 0.000 0.000 0.000 0.000 0.001 0.002 0.000 0.000 . N 102 102 102 102 90 93 70 60 102 10. Programming r 0.484** 0.526** 0.601** 0.322** 0.445** 0.356** 0.536** 0.506** 0.497** 1 Sig. 0.000 0.000 0.000 0.010 0.001 0.006 0.000 0.000 0.000 . N 64 64 64 64 53 59 52 49 55 64 11. 1st Semester GPA r 0.881** 0.906** 0.922** 0.533** 0.369** 0.445** 0.720** 0.695** 0.589** 0.612** 1 Sig. 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 . N 123 123 123 123 105 110 78 67 102 64 123 12. 2nd Semester GPA r 0.412** 0.390** 0.550** 0.298** 0.460** 0.087 0.659** 0.676** 0.656** 0.684** 0.499** 1 Sig. 0.000 0.000 0.000 0.001 0.000 0.378 0.000 0.000 0.000 0.000 0.000 . N 115 115 115 115 102 105 78 67 102 64 115 115 13. Cumulative GPA r 0.740** 0.727** 0.846** 0.451** 0.420** 0.271** 0.791** 0.757** 0.741** 0.707** 0.854** 0.862** 1 Sig. 0.000 0.000 0.000 0.000 0.000 0.005 0.000 0.000 0.000 0.000 0.000 0.000 . N 115 115 115 115 102 105 78 67 102 64 115 115 115 14. Corrects in GEFT-C r 0.180* 0.174 0.260** 0.174 0.112 0.050 0.157 0.115 0.125 0.182 0.217* 0.156 0.208* 1 Sig. 0.048 0.056 0.004 0.056 0.256 0.607 0.174 0.357 0.214 0.149 0.017 0.098 0.027 . N 121 121 121 121 104 108 77 66 101 64 121 114 114 123 **. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed). There were significant differences between the two FDI groups of students in Calculus I, Linear Algebra, 1st Semester GPA and Cumulative GPA. In Calculus I, students located in the half that is more FI performed significantly higher (M=4.5, SD=1.0) than did students located in the half that is less FI (M=4.1, SD=0.9); t(121)=2.56, p = 0.012. Also, in Linear Algebra, students located in the half that is more FI performed significantly higher (M=3.8, SD=1.1) than PREDICTING ENGINEERING STUDENT’S PERFORMANCE 52 did students located in the half that is less FI (M=3.2, SD=1.1); t(121)=3.17, p = 0.002. In addition, students located in the half that is more FI obtained a 1st Semester GPA that was significantly higher than that of students located in the half that is less FI (M=4.6, SD=0.8) (M=4.3, SD=0.7), t(121)=2.35, p = 0.021. Finally, Cumulative GPA was also significantly higher for students located in the more FI half (M=4.8, SD=0.6) (M=4.6, SD=0.6), t(113)=2.33, p = 0.022. These results may suggest that FDI has a more predictive value for the more mathematics-intense courses. Not finding a significant relationship between FDI and Calculus II, also a mathematics-intense course, could be explained because of the smaller sample size available for this course (76 students as compared to 121 in Calculus I and Linear Algebra), which has an impact on the statistical power to detect a significant difference. Table 8 Group Statistics for Independent Samples Test Between Students Who Scored Above and Below the Mean in the GEFT-C (For Field Dependence-Independence) GEFT-C N Mean Std. Deviation Std. Error Mean Calculus I Below the Mean 48 4.058 0.948 0.137 Above the Mean 75 4.524 1.005 0.116 Chemistry II Below the Mean 48 4.092 1.168 0.169 Above the Mean 75 4.508 1.305 0.151 Linear Algebra Below the Mean 48 3.179 1.094 0.158 Above the Mean 75 3.820 1.094 0.126 Challenges Engineering Below the Mean 48 4.983 0.467 0.068 Above the Mean 75 5.080 0.595 0.069 Communication Skills Below the Mean 39 5.433 0.747 0.120 Above the Mean 66 5.391 1.027 0.126 Ethics Below the Mean 43 5.319 0.672 0.103 Above the Mean 67 5.437 0.744 0.091 Calculus II Below the Mean 24 4.675 0.802 0.164 Above the Mean 54 4.913 0.738 0.101 Statics & Dynamics Below the Mean 19 3.842 1.413 0.324 Above the Mean 48 4.292 1.116 0.161 Economics Below the Mean 40 5.175 0.721 0.114 Above the Mean 62 5.335 0.728 0.093 Programming Below the Mean 18 4.778 1.145 0.270 Above the Mean 46 5.120 1.057 0.156 GPA 1 st Semester Below the Mean 48 4.299 0.725 0.105 Above the Mean 75 4.630 0.787 0.091 GPA 2 nd Semester Below the Mean 44 4.725 0.743 0.112 Above the Mean 71 4.933 0.673 0.080 Cumulative GPA Below the Mean 44 4.558 0.616 0.093 Above the Mean 71 4.823 0.579 0.069 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 53 Table 9 Independent Samples Test Between Students Who Scored Above and Below the Mean in the GEFT-C (For Field Dependence-Independence) Levene's Test for Equality of Variances t-test for Equality of Means F Sig. t df Sig. (2- tailed) Mean Diff. Std. Error Diff. 95% Conf. Interval of the Difference Lower Upper Calculus I Equal variances assumed 0.005 0.944 -2.562 121 0.012 -0.466 0.182 -0.826 -0.106 Equal variances not assumed -2.595 104.514 0.011 -0.466 0.179 -0.822 -0.110 Chemistry II Equal variances assumed 0.132 0.717 -1.797 121 0.075 -0.416 0.232 -0.875 0.042 Equal variances not assumed -1.841 108.213 0.068 -0.416 0.226 -0.865 0.032 Linear Algebra Equal variances assumed 0.002 0.969 -3.170 121 0.002 -0.641 0.202 -1.041 -0.241 Equal variances not assumed -3.170 100.300 0.002 -0.641 0.202 -1.042 -0.240 Challenges Engineering Equal variances assumed 0.307 0.581 -0.952 121 0.343 -0.097 0.102 -0.298 0.104 Equal variances not assumed -1.004 115.902 0.318 -0.097 0.096 -0.287 0.094 Communicati on Skills Equal variances assumed 0.814 0.369 0.225 103 0.822 0.042 0.189 -0.332 0.416 Equal variances not assumed 0.244 98.444 0.808 0.042 0.174 -0.303 0.388 Ethics Equal variances assumed 0.000 0.993 -0.847 108 0.399 -0.119 0.140 -0.397 0.159 Equal variances not assumed -0.866 96.230 0.388 -0.119 0.137 -0.391 0.153 Calculus II Equal variances assumed 0.220 0.641 -1.280 76 0.205 -0.238 0.186 -0.608 0.132 Equal variances not assumed -1.239 41.066 0.222 -0.238 0.192 -0.626 0.150 Statics & Dynamics Equal variances assumed 2.491 0.119 -1.376 65 0.174 -0.450 0.327 -1.102 0.203 Equal variances not assumed -1.242 27.349 0.225 -0.450 0.362 -1.192 0.293 Economics Equal variances assumed 1.720 0.193 -1.091 100 0.278 -0.161 0.147 -0.452 0.131 Equal variances not assumed -1.093 83.965 0.277 -0.161 0.147 -0.452 0.131 Programming Equal variances assumed 0.595 0.443 -1.137 62 0.26 -0.342 0.301 -0.943 0.259 Equal variances not assumed -1.097 29.013 0.282 -0.342 0.312 -0.979 0.295 GPA 1 st Semester Equal variances assumed 0.115 0.735 -2.348 121 0.021 -0.331 0.141 -0.611 -0.052 Equal variances not assumed -2.390 106.245 0.019 -0.331 0.139 -0.606 -0.057 GPA 2 nd Semester Equal variances assumed 0.612 0.436 -1.546 113 0.125 -0.208 0.134 -0.474 0.059 Equal variances not assumed -1.510 84.427 0.135 -0.208 0.138 -0.481 0.066 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 54 Table 9, continued Levene's Test for Equality of Variances t-test for Equality of Means F Sig. t df Sig. (2- tailed) Mean Diff. Std. Error Diff. 95% Conf. Interval of the Difference Lower Upper Cumulative GPA Equal variances assumed 0.022 0.883 -2.331 113 0.022 -0.265 0.114 -0.491 -0.040 Equal variances not assumed -2.297 86.937 0.024 -0.265 0.116 -0.495 -0.036 In addition to an independent-samples t-test and to understand whether FDI may also predict passing or failing a class, logistic regression was utilized. The four dependent variables that had significant Spearman's rho correlation coefficient where recoded into dummy variables. Students who obtained the minimum grade to pass (≥3.95) where coded as 1 and students who did not obtain the minimum grade to pass were coded as 0. The variable FDI was recoded also as a dummy variable where students who scored above the mean (M=21.54) in the GEFT-C were coded as 1 (22 and above) and students who scored below the mean were coded as 0 (21 and below). For each case, two models were analyzed, one without any control variable (Model 1) and another with PSU scores as a control variable (Model 2). The logistic regression results for the models with no control variables show a significant relationship between being a student from the more FI half and passing Calculus I (p=0.039) and Linear Algebra (p=0.006). However, no significant relationship was found between being above or below the mean in the FDI spectrum and obtaining a 1st Semester GPA or 1st Year Cumulative GPA equal or above 3.95. For students who fall in the more FI half, the odds of passing Calculus I, which means obtaining a grade 3.95 or higher, increased by 127% (p=0.039). At the same time, for the same students, the odds of passing Linear Algebra increased by 190% (p=0.008). PREDICTING ENGINEERING STUDENT’S PERFORMANCE 55 Interestingly, when controlling for PSU scores, FDI is not a significant predictor in any of the dependent variables. These results suggest that FDI might be a good predictor for the more mathematics-intense classes where the PSU is already an excellent and even better predictor. In addition, another interesting result is that, even though PSU scores are an excellent predictor of passing or failing in Calculus I, Linear Algebra and of 1st Semester GPA, they do not predict obtaining a 1st Year Cumulative GPA of 3.95 or higher, controlling for the other variables in the model. A summary of the logistic regression analysis results is presented in Tables 10, 11, 12 and 13. Table 10 Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in Calculus I Calculus I (N=121) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant 0.262 0.297 1.300 0.970* 0.379 2.638 Categorical GEFT-C 0.819* 0.399 2.267 0.042 0.475 1.043 PSU 1.349*** 0.295 3.854 χ 2 4.242 33.118 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Table 11 Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in Linear Algebra Linear Algebra (N=121) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant -0.932** 0.327 0.394 -0.538 0.387 0.584 Categorical GEFT-C 1.065** 0.401 2.901 0.205 0.493 1.227 PSU 1.632*** 0.323 5.112 χ 2 7.462 47.952 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 56 Table 12 Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st Semester GPA 1st Semester GPA (N=121) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant 0.932** 0.327 2.538 1.481*** 0.403 4.399 Categorical GEFT-C 0.727 0.454 2.068 0.152 0.502 1.164 PSU 0.925** 0.286 2.522 χ 2 2.555 14.710 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Table 13 Summary of Logistic Regression Analysis for Field Dependence-Independence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st year Cumulative GPA Cumulative GPA (N=114) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant 2.028*** 0.476 7.600 2.315*** 0.552 10.127 Categorical GEFT-C 0.790 0.701 2.204 0.452 0.742 1.572 PSU 0.517 0.394 1.676 χ 2 1.278 3.063 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Logical-Mathematical Intelligence and Students Performance LMI, measured through the TILS, had a negatively skewed bimodal distribution. Some explorations were undertaken to understand if the bimodal distribution was due to differences between “Talent” students above the PSU cut score, “Talent” students below the PSU cut score and “Regular Admission” students. However, all of them kept the bimodal condition when analyzed separately. Conservatively, Spearman's rho correlation coefficients were calculated, which are reported in Table 14. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 57 The Relationship Between Logical-Mathematical Intelligence and Student Performance LMI has a positive, small, but significant correlation with Calculus I, r(118)=0.208, p=0.022; Linear Algebra, r(118)=0.232, p=0.011; 1st Semester GPA, r(118)=0.207, p=0.023; and Cumulative GPA r(111)=0.196, p=0.038. It is worth mentioning that, even when significant, these correlations are very small and they are hard to see from observing scatterplots between LMI and all dependent variables. Logical-Mathematical Intelligence as a Predictor of Students Performance To evaluate whether LMI may predict students’ grades, an independent-samples t-test was conducted to compare grades between students who where above and below the average in LMI for all the dependent variables. The results, reported in Tables 15 and 16, show significant differences between the two groups of students in Linear Algebra, Statics and Dynamics, and cumulative GPA. In Linear Algebra, students who were above the average in LMI performed significantly higher (M=3.8, SD=1.2) than did students who were below the average in LMI (M=3.3, SD=1.0); t(118)=2.27, p = 0.025. In addition, students who were above the average in LMI also performed significantly better in Statics and Dynamics (M=4.5, SD=1.0) than did students below the average in LMI (M=3.7, SD=1.4); t(32)=2.39, p = 0.023. Finally, for the same two groups of students, the ones above the average obtained a slightly but significantly better 1st year cumulative GPA (M=4.8, SD=0.6) than did the ones below the average (M=4.6, SD=0.6); t(111)=2.05, p = 0.043. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 58 Table 14 Spearman's rho Correlations Coefficients for Logical-Mathematical Intelligence (measured through the TILS) and DV’s Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1. Calculus I r 1 Sig. . N 123 2. Chemistry II r 0.767** 1 Sig. 0.000 . N 123 123 3. Linear Algebra r 0.827** 0.819** 1 Sig. 0.000 0.000 . N 123 123 123 4. Challenges Engineering r 0.282** 0.441** 0.391** 1 Sig. 0.002 0.000 0.000 . N 123 123 123 123 5. Communication Skills r 0.319** 0.177 0.305** 0.204* 1 Sig. 0.001 0.071 0.002 0.037 . N 105 105 105 105 105 6. Ethics r 0.297** 0.307** 0.285** 0.376** 0.155 1 Sig. 0.002 0.001 0.003 0.000 0.141 . N 110 110 110 110 92 110 7. Calculus II r 0.662** 0.600** 0.734** 0.318** 0.138 0.356** 0.003 69 1 Sig. 0.000 0.000 0.000 0.005 0.258 . N 78 78 78 78 69 78 8. Statics & Dynamics r 0.656** 0.678** 0.736** 0.325** -0.013 0.194 0.676** 1 Sig. 0.000 0.000 0.000 0.007 0.925 0.135 0.000 . N 67 67 67 67 58 61 67 67 9. Economics r 0.496** 0.468** 0.536** 0.476** 0.344** 0.320** 0.575** 0.546** 1 Sig. 0.000 0.000 0.000 0.000 0.001 0.002 0.000 0.000 . N 102 102 102 102 90 93 70 60 102 10. Programming r 0.484** 0.526** 0.601** 0.322** 0.445** 0.356** 0.536** 0.506** 0.497** 1 Sig. 0.000 0.000 0.000 0.010 0.001 0.006 0.000 0.000 0.000 . N 64 64 64 64 53 59 52 49 55 64 11. 1st Semester GPA r 0.881** 0.906** 0.922** 0.533** 0.369** 0.445** 0.720** 0.695** 0.589** 0.612** 1 Sig. 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 . N 123 123 123 123 105 110 78 67 102 64 123 12. 2nd Semester GPA r 0.412** 0.390** 0.550** 0.298** 0.460** 0.087 0.659** 0.676** 0.656** 0.684** 0.499** 1 Sig. 0.000 0.000 0.000 0.001 0.000 0.378 0.000 0.000 0.000 0.000 0.000 . N 115 115 115 115 102 105 78 67 102 64 115 115 13. Cumulative GPA r 0.740** 0.727** 0.846** 0.451** 0.420** 0.271** 0.791** 0.757** 0.741** 0.707** .854** 0.862** 1 Sig. 0.000 0.000 0.000 0.000 0.000 0.005 0.000 0.000 0.000 0.000 0.000 0.000 . N 115 115 115 115 102 105 78 67 102 64 115 115 115 14. Corrects in TILS r 0.208* 0.143 0.232* 0.102 -0.067 0.039 0.139 0.140 0.086 0.084 0.207* 0.084 0.196* 1 Sig. 0.022 0.118 0.011 0.265 0.500 0.691 0.232 0.265 0.392 0.515 0.023 0.378 0.038 . N 120 120 120 120 103 108 76 65 100 63 120 113 113 122 **. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed). PREDICTING ENGINEERING STUDENT’S PERFORMANCE 59 Table 15 Group Statistics for Independent Samples Test Between Students Who Scored Above and Below the Mean in the TILS (for Logical-Mathematical Intelligence) TILS N Mean Std. Deviation Std. Error Mean Calculus I Below the Mean 53 4.185 0.879 0.121 Above the Mean 67 4.470 1.063 0.130 Chemistry II Below the Mean 53 4.200 1.081 0.149 Above the Mean 67 4.473 1.375 0.168 Linear Algebra Below the Mean 53 3.313 0.997 0.137 Above the Mean 67 3.778 1.197 0.146 Challenges Engineering Below the Mean 53 4.996 0.440 0.060 Above the Mean 67 5.084 0.632 0.077 Communication Skills Below the Mean 46 5.535 0.696 0.103 Above the Mean 57 5.323 1.058 0.140 Ethics Below the Mean 46 5.446 0.636 0.094 Above the Mean 62 5.374 0.754 0.096 Calculus II Below the Mean 28 4.618 0.773 0.146 Above the Mean 48 4.958 0.730 0.105 Statics & Dynamics Below the Mean 22 3.682 1.378 0.294 Above the Mean 43 4.465 0.956 0.146 Economics Below the Mean 46 5.185 0.754 0.111 Above the Mean 54 5.346 0.713 0.097 Programming Below the Mean 17 4.788 1.243 0.302 Above the Mean 46 5.122 1.029 0.152 GPA 1 st Semester Below the Mean 53 4.393 0.623 0.086 Above the Mean 67 4.600 0.867 0.106 GPA 2 nd Semester Below the Mean 51 4.762 0.731 0.102 Above the Mean 62 4.935 0.685 0.087 Cumulative GPA Below the Mean 51 4.594 0.589 0.083 Above the Mean 62 4.826 0.610 0.078 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 60 Table 16 Independent Samples Test for Students Who Scored Above and Below the Mean in the TILS (for Logical-Mathematical Intelligence) Levene's Test for Equality of Variances F Sig. t df Sig. (2- tailed) Mean Diff. Std. Error Diff. 95% Conf. Interval of the Difference Lower Upper Calculus I Equal variances assumed 1.639 0.203 -1.574 118 0.118 -0.285 0.181 -0.644 0.074 Equal variances not assumed -1.609 117.745 0.110 -0.285 0.177 -0.636 0.066 Chemistry II Equal variances assumed 2.264 0.135 -1.185 118 0.239 -0.273 0.231 -0.730 0.184 Equal variances not assumed -1.218 117.998 0.226 -0.273 0.224 -0.717 0.171 Linear Algebra Equal variances assumed 2.555 0.113 -2.269 118 0.025 -0.464 0.205 -0.870 -0.059 Equal variances not assumed -2.318 117.655 0.022 -0.464 0.200 -0.861 -0.068 Challenges Engineering Equal variances assumed 1.232 0.269 -0.855 118 0.394 -0.087 0.102 -0.290 0.115 Equal variances not assumed -0.891 116.254 0.375 -0.087 0.098 -0.282 0.107 Communication Skills Equal variances assumed 1.599 0.209 1.170 101 0.245 0.212 0.181 -0.147 0.571 Equal variances not assumed 1.221 97.323 0.225 0.212 0.174 -0.133 0.557 Ethics Equal variances assumed 0.612 0.436 0.520 106 0.604 0.072 0.137 -0.201 0.344 Equal variances not assumed 0.533 104.219 0.595 0.072 0.134 -0.194 0.337 Calculus II Equal variances assumed 0.053 0.818 -1.920 74 0.059 -0.341 0.177 -0.694 0.013 Equal variances not assumed -1.891 53.994 0.064 -0.341 0.180 -0.701 0.021 Statics & Dynamics Equal variances assumed 8.978 0.004 -2.680 63 0.009 -0.783 0.292 -1.367 -0.199 Equal variances not assumed -2.388 31.661 0.023 -0.783 0.328 -1.452 -0.115 Economics Equal variances assumed 0.268 0.606 -1.100 98 0.274 -0.162 0.147 -0.453 0.130 Equal variances not assumed -1.095 93.566 0.276 -0.162 0.148 -0.455 0.131 Programming Equal variances assumed 1.483 0.228 -1.079 61 0.285 -0.334 0.309 -0.952 0.285 Equal variances not assumed -0.988 24.574 0.333 -0.334 0.338 -1.029 0.362 GPA 1 st Semester Equal variances assumed 7.343 0.008 -1.463 118 0.146 -0.207 0.141 -0.487 0.073 Equal variances not assumed -1.52 116.995 0.131 -0.207 0.136 -0.477 0.063 GPA 2 nd Semester Equal variances assumed 0.003 0.956 -1.295 111 0.198 -0.173 0.134 -0.437 0.092 Equal variances not assumed -1.286 103.911 0.201 -0.173 0.134 -0.439 0.094 Cumulative GPA Equal variances assumed 0.729 0.395 -2.046 111 0.043 -0.232 0.114 -0.457 -0.007 Equal variances not assumed -2.053 108.146 0.042 -0.232 0.113 -0.457 -0.008 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 61 To understand whether LMI may also predict passing or failing a class, logistic regression was utilized. Calculus I, Linear Algebra, Statics and Dynamics, 1st Semester GPA and cumulative GPA were recoded into dummy variables. These variables represent the four dependent variables that had significant Spearman's rho correlation coefficient plus Statics and Dynamics, which showed significant differences for the independent sample t-test. Students who obtained the minimum grade to pass (≥3.95) were coded as 1 and students who did not obtain the minimum grade to pass were coded as 0. For each case, two models were analyzed, one without control variables (Model 1) and another with PSU scores as a control variable (Model 2). A summary of the results is reported in Tables 17, 18, 19, 20 and 21. The results for Model 1 show a significant relationship between students’ LMI and passing Calculus I (p=0.017) and Linear Algebra (p=0.011). However, no significant relationship was found between students’ LMI and passing Statics and Dynamics and having a 1st Semester GPA or a 1st year Cumulative GPA equal or above 3.95. In the cases of Calculus I and Linear Algebra, for each 1 SD increase in students’ LMI, the odds of passing the class, which means obtaining a grade 3.95 or above, increased by 63% (p=0.017) and 78% (p=0.006), respectively. Interestingly, when controlling for PSU scores (Model 2), LMI is not a significant predictor of any of the dependent variables. These results suggest that LMI may be a good predictor for the more mathematics-intense classes, where the PSU is already an excellent and even a better predictor. In addition, and similar to the FDI results, another interesting finding is that, even though PSU scores are an excellent predictor for passing or failing Calculus I, Linear Algebra, Statics and Dynamics, and having a 1st Semester GPA of 3.95 or higher, it does not predict obtaining a 1st year Cumulative GPA of 3.95 or higher, controlling for the other variables in the model. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 62 Table 17 Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in Calculus I Calculus I (N=120) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant 0.751*** 0.201 2.119 0.975*** 0.245 2.650 TILS 0.491* 0.206 1.634 0.088 0.242 1.092 PSU 1.316*** 0.296 3.728 χ 2 5.957 33.199 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Table 18 Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in Linear Algebra Linear Algebra (N=120) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant -0.314 0.193 0.731 -0.461 0.237 0.631 TILS 0.578** 0.212 1.783 0.151 0.254 1.163 PSU 1.638*** 0.324 5.145 χ 2 8.25 48.623 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Table 19 Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in Statics and Dynamics Statics and Dynamics (N=65) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant 0.518 0.271 1.678 0.043 0.334 1.044 TILS 0.334 0.301 1.397 -0.002 0.344 0.998 PSU 1.436** 0.426 4.206 χ 2 1.252 17.098 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 63 Table 20 Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st Semester GPA 1st Semester GPA (N=120) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant 1.334*** 0.225 3.797 1.612*** 0.280 5.012 TILS 0.042 0.229 1.043 -0.393 0.278 0.675 PSU 1.126*** 0.311 3.083 χ 2 0.034 16.699 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Table 21 Summary of Logistic Regression Analysis for Logical-Mathematical Intelligence as a Predictor of Obtaining a Grade 3.95 or Higher in 1 st Year Cumulative GPA Cumulative GPA (N=113) Model 1 Model 2 Predictor B SE B e B B SE B e B Constant 2.461*** 0.352 11.719 2.563*** 0.380 12.974 TILS 0.209 0.340 1.232 -0.010 0.380 0.990 PSU 0.590 0.400 1.805 χ 2 0.367 2.684 df 1 2 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Task Persistence and Students Performance TP was measured with the TIN, which, in contrast to the tests used to measure FDI and LMI, has not been broadly validated. Similar to what French (1948) did in his work with freshman engineering students, in this study the relationship between several measures from the TIN and freshman engineering student performance also was examined. Six variables were computed from the TIN: number of items answered correctly, number of items answered with an X, highest time-number, number of items answered during the 2nd half hour, student guessing and total number of reversals. Correlations between these variables and all the dependent variables were analyzed, and, because most of the TIN variables are not normally distributed, the PREDICTING ENGINEERING STUDENT’S PERFORMANCE 64 Spearman's rho correlation coefficient was calculated to explored potential relationships among variables (refer to Table 22). The Relationship Between Task Persistence and Students Performance Among the six TIN variables, the variable “number of items answered correctly” showed the highest correlation with grades among several classes, 1st semester GPA and cumulative GPA. The “number of items answered correctly” has a medium, positive correlation with Statics and Dynamics, r=0.33, n=67, p=0.007 and Programming, r=0.33, n=64, p=0.009. It also has a small, positive correlation with Linear Algebra, r=0.28, n=123, p=0.002; Calculus I, r=0.21, n=123, p=0.017; Chemistry, r=0.19, n=123, p=0.039; 1st Semester GPA, r=0.21, n=123, p=0.018; and Cumulative GPA, r=0.22, n=115, p=0.016. The variable “number of items answered with an X” had, as expected, a negative but small correlation with Chemistry, r=-0.22, n=123, p=0.013; however, had no significant correlation with the rest of the dependent variables. The previous results must be interpreted with caution because of several reasons. First, the “number of items answered correctly” is probably measuring both ability and persistence, and, therefore, a direct association with persistence it is not appropriate. Second, even though it was expected that “number of items answered with an X” would be a more direct measure of persistence, this variable had only one significant correlation among the 13 dependent variables. Finally, and contrary to what was expected, the variables “number of items answered correctly” and “number of items answered with an X” had no correlation, r=-0.07, which means they measure different things. A possible interpretation for this is that the variable “number of items answered with an X” is measuring persistence while “number of items answered correctly” is not. However, this option seems unlikely given that the TIN proved to be difficult enough for all students, having the maximum score of 25 over the 30 questions with a mean of 16.01, meaning PREDICTING ENGINEERING STUDENT’S PERFORMANCE 65 that effort and TP are probably being measured by the “number of items answered correctly.” If this is the case, the variable “number of items answered with an X” it is not a good measurement of persistence. This may be because students answered with X’s the questions they really thought did not have a solution and did not necessarily give up on those questions. Many of the questions in the TIN were highly difficult, and, therefore, it is likely that students may have thought those problems did not have a solution. It could be also that none of them are good measurements of TP. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 66 Table 22 Spearman's rho Correlations Coefficients for Task Persistence and DV’s Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 1. Calculus I r 1 Sig. . N 123 2. Chemistry II r .767** 1 Sig. .000 . N 123 123 3. Linear Algebra r .827** .819** 1 Sig. .000 .000 . N 123 123 123 4. Challenge Enginee- ring r .282** .441** .391** 1 Sig. .002 .000 .000 . N 123 123 123 123 5. Commu- nication Skills r .319** .177 .305** .204* 1 Sig. .001 .071 .002 .037 . N 105 105 105 105 105 6. Ethics r .297** .307** .285** .376** .155 1 Sig. .002 .001 .003 .000 .141 . N 110 110 110 110 92 110 7. Calculus II r .662** .600** .734** .318** .138 .356** 1 Sig. .000 .000 .000 .005 .258 .003 . N 78 78 78 78 69 69 78 8. Statics & Dynamics r .656** .678** .736** .325** -.013 .194 .676** 1 Sig. .000 .000 .000 .007 .925 .135 .000 . N 67 67 67 67 58 61 67 67 9. Eco- nomics r .496** .468** .536** .476** .344** .320** .575** .546** 1 Sig. .000 .000 .000 .000 .001 .002 .000 .000 . N 102 102 102 102 90 93 70 60 102 1. Program- ming r .484** .526** .601** .322** .445** .356** .536** .506** .497** 1 Sig. .000 .000 .000 .010 .001 .006 .000 .000 .000 . N 64 64 64 64 53 59 52 49 55 64 11. 1st Semester GPA r .881** .906** .922** .533** .369** .445** .720** .695** .589** .612** 1 Sig. .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 . N 123 123 123 123 105 110 78 67 102 64 123 12. 2nd Semester GPA r .412** .390** .550** .298** .460** .087 .659** .676** .656** .684** .499** 1 Sig. .000 .000 .000 .001 .000 .378 .000 .000 .000 .000 .000 . N 115 115 115 115 102 105 78 67 102 64 115 115 13. Cum. GPA r .740** .727** .846** .451** .420** .271** .791** .757** .741** .707** .854** .862** 1 Sig. .000 .000 .000 .000 .000 .005 .000 .000 .000 .000 .000 .000 . N 115 115 115 115 102 105 78 67 102 64 115 115 115 14. Corrects in TIN r .214* .186* .281** .080 .002 .065 .191 .326** .143 .326** .213* .158 .224* 1 Sig. .017 .039 .002 .381 .987 .503 .093 .007 .152 .009 .018 .092 .016 . N 123 123 123 123 105 110 78 67 102 64 123 115 115 125 15. Mark with X in TIN r -.151 -.223* -.135 -.059 -.118 .020 .078 .020 .003 .037 -.170 .052 -.041 -.072 1 Sig. .095 .013 .136 .516 .230 .835 .495 .871 .979 .769 .061 .584 .666 .428 . N 123 123 123 123 105 110 78 67 102 64 123 115 115 125 125 16. Reversals on TIN r .103 .084 .021 .050 .051 .030 .059 .156 .004 .170 .062 -.047 .015 -.041 .094 1 Sig. .256 .354 .815 .585 .608 .752 .606 .208 .966 .180 .494 .617 .876 .646 .300 . N 123 123 123 123 105 110 78 67 102 64 123 115 115 125 125 125 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 67 Table 22, continued Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 17. Max time in the TIN r -.173 -.178* -.241** .050 .020 .020 -.246* -.238 -.114 .019 -.144 -.102 -.162 .062 .150 -.033 1 Sig. .056 .049 .008 .581 .838 .840 .030 .052 .258 .884 .112 .282 .084 .496 .096 .715 . N 122 122 122 122 104 109 78 67 101 64 122 114 114 124 124 124 124 18. Questions answered in the 2nd 1/2 hour in TIN r -.217* -.215* -.193* -.036 .033 -.062 -.260* -.190 -.034 .019 -.187* -.015 -.128 .060 .401** -.006 .571** 1 Sig. .016 .017 .033 .693 .740 .524 .021 .125 .737 .879 .039 .872 .173 .509 .000 .943 .000 . N 122 122 122 122 104 109 78 67 101 64 122 114 114 124 124 124 124 124 19. Student Guessing in TIN r -.095 -.164 -.074 .015 -.053 .047 -.023 -.021 .054 .057 -.095 .049 -.011 .113 .723** .060 .227* .532** 1 Sig. .297 .070 .413 .872 .592 .625 .844 .863 .587 .657 .295 .604 .903 .211 .000 .503 .011 .000 . N 123 123 123 123 105 110 78 67 102 64 123 115 115 125 125 125 124 124 125 ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). Two variables that negatively correlate with performance are the “highest time-number” and the “number of items answered during the 2nd half hour.” Both showed small, negative correlations with several courses and 1st semester GPA. The variable “highest time-number” had significant correlations with Chemistry, r=-0.19, n=122, p=0.049; Linear Algebra, r=-0.24, n=122, p=0.008; and Calculus II, r=-0.25, n=78, p=0.030. The variable “number of items answered during the 2nd half hour” had significant correlations with Calculus I, r=-0.22, n=122, p=0.016; Chemistry, r=-0.22, n=122, p=0.017; Linear Algebra, r=-0.19, n=122, p=0.033; Calculus II, r=-0.26, n=78, p=0.021; and 1 st semester GPA, r=-0.19, n=122, p=0.039. Interestingly, theses two independent variables have a large, positive correlation, r=0.57, n=124, p<0.01, even though intuitively, the maximum time scored on the test (time when the student marked the last answer) has nothing to do with how many questions the student answered during the second half hour of the test. Perhaps, the student, during the second half hour, under the pressure of time, started to “guess” some of the answers. The previous is just one possible interpretation, but this might also be explained by, first, the negative relationship between these two variables and performance; second, the fact that these two variables have a large, positive correlation between them, r=0.57, n=124, p<0.01; and, third, the fact that the variables “number PREDICTING ENGINEERING STUDENT’S PERFORMANCE 68 of items answered during the 2nd half hour” and “student guessing” also have a large, positive correlation, r=0.52, n=124, p<0.01. The variables “student guessing” and “total number of reversals” did not show significant relationship with any of the performance measures. Task Persistence as a Predictor of Students Performance To evaluate whether TP may predict students’ grades, multiple regression was conducted to analyze each dependent variable where more than one TIN variable showed to have significant Spearman's rho correlation coefficient. Given the high correlation between the variables “highest time-number” and “number of items answered during the 2nd half hour,” for the cases where both of them had significant Spearman's rho correlation coefficient, only the variable that contributed more to the model was included in the equation. For each case, two models were analyzed, one with only the TIN variables and another with the TIN variables and PSU scores as a control variable. Considering the previous, the following multiple regression equations were analyzed: • Calculus I: Model 1: β 0 + β 1 *zTINcorrect + β 2 *zTIN2nd_half_hour Model 2: β 0 + β 1 *zTINcorrect + β 2 *zTIN2nd_half_hour + β 3 *zPSU • Chemistry II: Model 1: β 0 + β 1 *zTINcorrect + β 2 *zTINisX + β 3 *zTIN2nd_half_hour Model 2: β 0 + β 1 *zTINcorrect + β 2 *zTINisX + β 3 *zTIN2nd_half_hour + β 4 *zPSU • Linear Algebra: Model 1: β 0 + β 1 *zTINcorrect + β 2 *zTIN2nd_half_hour Model 2: β 0 + β 1 *zTINcorrect + β 2 *zTIN2nd_half_hour + β 3 *zPSU PREDICTING ENGINEERING STUDENT’S PERFORMANCE 69 • GPA1: Model 1: β 0 + β 1 *zTINcorrect + β 2 *zTIN2nd_half_hour Model 2: β 0 + β 1 *zTINcorrect + β 2 *zTIN2nd_half_hour + β 3 *zPSU The assumptions for linear regression were checked for each model. It was verified that residuals normally distributed about the predicted DV scores, that residuals have a straight-line relationship with predicted DV scores, and finally that the variance of the residuals about the predicted DV scores are the same for all predicted scores. Visual inspection of Normal P-P plots and scatter plots of the standardized residuals show there were not concerning problems regarding the previous assumptions. A summary of the regression analysis results for models 1 and 2 is presented in Tables 23, 24, 25 and 26. As is shown in the equation for Calculus I – Model 1, two TIN variables had significant Spearman's rho correlation coefficient: the number of items answered correctly and the number of items answered during the 2nd half hour. The results suggest that the model explains 10% of the variance in Calculus I grades (p=0.002). The average final grade in Calculus I for students at the mean in the number of items answered correctly (16.01) and the number of items answered during the 2nd half hour (7.19) was 4.34 (p<0.001). Controlling for the other variables in the model, a 1 SD increase in the number of items answered correctly in the TIN is associated with 0.22 points increase in the student’s grade (p=0.012). In addition, a 1 SD increase in the number of items answered during the 2nd half hour in the TIN is associated with 0.23 points decrease in the student’s grade (p=0.010). For the case of Chemistry II – Model 1, four TIN variables had significant Spearman's rho correlation coefficients: the number of items answered correctly, the number of items answered with an X, the highest time-number and the number of items answered during the 2nd PREDICTING ENGINEERING STUDENT’S PERFORMANCE 70 half hour. The variable highest time-number was not included in the model because did not contribute to increase R 2 (amount of variance explained by the model) and was highly correlated with the variable number of items answered during the 2nd half hour. The results showed that, even though the model may explain 7% of the variance in Chemistry II grades (p=0.030), none of the TIN variables are good predictors of student performance in Chemistry II. This is because the independent variables were highly correlated. In the case of Linear Algebra – Model 1, three TIN variables had significant Spearman's rho correlation coefficients: the number of items answered correctly, the highest time-number and the number of items answered during the 2nd half hour. Similar to the case of Chemistry II, the variable highest time-number was not included in the model because it did not contribute to increase R 2 and was highly correlated with the variable number of items answered during the 2nd half hour. The results showed that the model explains 12% of the variance in Linear Algebra final grades (p=0.001). Specifically, the average final grade in Linear Algebra for students at the mean in the number of items answered correctly (16.01) and the number of items answered during the 2nd half hour (7.19) was 3.57 (p<0.001). In addition, a 1 SD increase in the number of items answered correctly in the TIN is associated with 0.32 points increase in the student’s grade (p=0.002). Also, a 1 SD increase in the number of items answered during the 2nd half hour in the TIN is associated with 0.25 points decrease in the student’s grade (p=0.013), controlling for the other variables in the model. Finally, for the case of 1st semester GPA – Model 1, two TIN variables had significant Spearman's rho correlation coefficient: the number of items answered correctly and the number of items answered during the 2nd half hour. The results show that the model only explains 7% of the variance for 1 st semester GPA (p=0.011). The average first semester GPA for students at the PREDICTING ENGINEERING STUDENT’S PERFORMANCE 71 mean in the number of items answered correctly (16.01) and the number of items answered during the 2nd half hour (7.19) was 4.50 (p<0.001). In particular, a 1 SD increase in the number of items answered correctly in the TIN is associated with 0.15 points increase in the student’s grade (p=0.035). In addition, a 1 SD increase in the number of items answered during the 2nd half hour in the TIN is associated with 0.16 points decrease in the student’s grade (p=0.023), controlling for the other variables in the model. The results for the TIN suggest that student performance for the more mathematics- intense courses can be predicted from two of the sub-dimensions of the test: the number of items answered correctly and the number of items answered during the 2nd half hour. However, as is seen in Tables 23 to 26, the PSU scores are a much better predictor than any of the sub- dimensions of the TIN, making insignificant any contribution of the TIN when controlling for PSU scores (Model 2 equations). This finding suggests that the TIN does not provide any additional information to the PSU in regards to grades prediction in Calculus I, Chemistry II, Linear Algebra and 1st Semester GPA. Table 23 Summary of Regression Analysis for Variables Predicting Student’s Performance in Calculus I Calculus I (N= 123) Model 1 Model 2 Predictor B SE (B) B SE (B) Intercept 4.342 *** 0.087 4.342 *** 0.072 Number of items answered correctly 0.224 * 0.088 -0.041 0.080 Number of items answered during the 2nd half hour in TIN -0.229 * 0.088 -0.042 0.076 PSU scores 0.634 *** 0.083 R 2 0.097 0.395 F 6.373 25.649 P 0.002 0.000 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 72 Table 24 Summary of Regression Analysis for Variables Predicting Student’s Performance in Chemistry II Chemistry II (N= 123) Model 1 Model 2 Predictor B SE (B) B SE (B) Intercept 4.346 *** 0.112 4.346 *** 0.092 Number of items answered correctly 0.167 0.113 -0.158 0.103 Number of items answered with an X -0.187 0.122 -0.028 0.103 Number of items answered during the 2nd half hour in TIN -0.166 0.122 0.014 0.103 PSU scores 0.819 *** 0.109 R 2 0.074 0.375 F 3.123 17.544 P 0.029 0.000 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 Table 25 Summary of Regression Analysis for Variables Predicting Student’s Performance in Linear Algebra Linear Algebra (N= 123) Model 1 Model 2 Predictor B SE (B) B SE (B) Intercept 3.570 *** 0.097 3.570 *** 0.078 Number of items answered correctly 0.317 ** 0.098 0.010 0.087 Number of items answered during the 2nd half hour in TIN -0.245 * 0.098 -0.029 0.083 PSU scores 0.733 *** 0.091 R 2 0.119 0.433 F 8.070 30.074 P 0.001 0.000 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 73 Table 26 Summary of Regression Analysis for Variables Predicting Student’s Performance for 1 st Semester GPA 1 st Semester GPA (N= 123) Model 1 Model 2 Predictor B SE (B) B SE (B) Intercept 4.501 *** 0.068 4.501 *** 0.055 Number of items answered correctly 0.147 * 0.069 -0.071 0.061 Number of items answered during the 2nd half hour in TIN -0.158 * 0.069 -0.005 0.058 PSU scores 0.520 *** 0.064 R 2 0.073 0.408 F 4.698 27.159 P 0.011 0.000 Note. Cell entries are estimated standardized regression coefficients for independent variables. *p<0.05, **p<0.01, ***p<0.001 PREDICTING ENGINEERING STUDENT’S PERFORMANCE 74 CHAPTER FIVE: DISCUSSION The main goal of this study was to evaluate FDI, LMI and TP as possible predictors of freshman engineering student performance. The motivation is rooted in the context of PUC’s T+I Program, which seeks to implement additional measures to orient the selection of students entering via this program. Still, the implications of findings may be much broader, serving to inform and guide other universities and programs in Chile. The following chapter presents the discussion of findings derived from the analysis and is organized by answering to the three research questions: 1. To what extent does Field Dependence-Independence (FDI) predict academic performance of engineering students during their first year of college? 2. To what extent does Logical-Mathematical Intelligence (LMI) predict academic performance of engineering students during their first year of college? 3. To what extent does Task Persistence (TP) predict academic performance of engineering students during their first year of college? Following the discussion of findings, are presented the implications for practice, propositions for future research and final conclusions. Discussion of Findings Field Dependence-Independence as a Predictor of Academic Performance of Engineering Students During Their First Year of College There were three major findings in regards to FDI. The first one is that FDI showed to be a significant predictor for student grades in the more mathematics-intense courses during the first semester. Students who were above the average on the FDI spectrum performed significantly higher in Calculus I (M=4.5, SD=1.0) and Linear Algebra (M=3.8, SD=1.1) than did students PREDICTING ENGINEERING STUDENT’S PERFORMANCE 75 below the average on the FDI spectrum (M=4.1, SD=0.9; M=3.2, SD=1.1). Related to the previous finding, significant differences were found for students above and below the average in FDI and their 1st semester GPA (M=4.6, SD=0.8; M=4.3, SD=0.7) and Cumulative GPA (M=4.8, SD=0.6; M=4.6, SD=0.6). A second major finding was that FDI might also predict passing or failing those courses. For students who are classified in the more FI half, the odds of passing Calculus I, which means obtaining a grade 3.95 or higher, increased by 127% (p=0.039). At the same time, for the same students, the odds of passing Linear Algebra increased by 190% (p=0.008). Finally, a third major finding was that, when controlling for the PSU), FDI is not a significant predictor of any of the dependent variables. This is especially interesting in the context of this study because the reason for implementing additional measures for selecting students entering via the T+I Program was to identify complementary tests to the PSU, which could provide additional information in regards to students’ skills. However, the results of this study show that FDI does not provide any additional information to what is already provided by the PSU. Logical-Mathematical Intelligence as a Predictor of Academic Performance of Engineering Students During Their First Year of College There were also three major findings for the case of LMI. The first important finding was that LMI showed to be a significant predictor of student grades for two 1st year classes: Linear Algebra (a class from the 1 st semester) and Statics and Dynamics (a class from the 2nd semester). Related to the previous, LMI was also a good predictor of Cumulative GPA. Students who were above the average in LMI performed significantly higher than did students who were below the average. For Linear Algebra, the difference between students who performed above versus below the average in LMI was 0.5 grade points (M=3.8, SD=1.2; M=3.3, SD=1.0). For Statics and PREDICTING ENGINEERING STUDENT’S PERFORMANCE 76 Dynamics, the difference between the two groups was 0.8 grade points (M=4.5, SD=1.0; M=3.7, SD=1.4), and, for Cumulative GPA, the difference between the two groups was 0.2 grade points (M=4.8, SD=0.6; M=4.6, SD=0.6). A second important finding was that LMI might predict passing or failing Linear Algebra and Calculus I. Interestingly, and unlike the case of FDI, even though LMI was a significant grade predictor for Statics and Dynamics and Cumulative GPA, it was neither a significant predictor of passing or failing Statics and Dynamics nor a significant predictor of having a Cumulative GPA that is equal or higher than 3.95. In the cases of Calculus I and Linear Algebra, for each 1 SD increase in students’ LMI, the odds of passing the class, which means obtaining a grade 3.95 or above, increased by 63% (p=0.017) for Calculus I and 78% (p=0.006) for Linear Algebra. Finally, a third important finding was that, when controlling for PSU scores, LMI is not a significant predictor for any of the dependent variables. Therefore, LMI does not provide any additional information compared to what the PSU already provides. This finding suggests that the LMI may be a good predictor for the more mathematics-intense classes, but the PSU is an even a better predictor. Task Persistence as a Predictor of Academic Performance of Engineering Students During Their First Year of College Before discussing the major findings in regards to TP, it is important to recapitulate the limitations concerning the instrument utilized to measure this variable. The first and most important limitation is that the TIN was not previously validated, and, therefore, the use of this test was exploratory. Cronbach’s Alpha results for the different sub-dimensions measured with the TIN suggest that more than one underlying concept is being measured by at least two dimensions: the number items that were answered correctly and the number of items answered during the 2nd half hour. Furthermore, these two sub-dimensions were the ones that best PREDICTING ENGINEERING STUDENT’S PERFORMANCE 77 predicted student performance from the six sub-dimensions measured by that test, making it difficult to draw specific conclusions from the results. In addition, the variable “number of items answered with an X” that is supposed to provide the most direct measure of TP was not a good predictor of student performance. The lack of relationship between this variable and student performance can be interpreted by accepting the lack of relationship between TP and performance, by distrusting the criterion validity of this sub- dimension from the TIN, or by justifying the lack of a finding to problems related to the design of this study such as a relatively small sample size. Since the literature broadly supports the relationship between TP and performance (Anderson & Bergman, 2011; Boe et al., 2002), the first interpretation is unlikely. In addition, French (1948), the author of the TIN test, did not find a significant relationship between the variable “number of items answered with an X” and students’ performance for engineering students, raising suspicions about the validity of the sub- dimension “number of items answered with an X” in measuring TP. Another interesting aspect in regards to the TIN is that the best sub-dimensions of the TIN in the prediction of student performance found in this study differ from French’s (1948) results. French (1948) found the highest positive correlations between students’ grades and the sub-dimensions “highest time-number” and “number of reversals” from the TIN. However, and contrary to French (1948), the results of this study suggest mostly negative correlations between “the highest time-number” and grades and no significant correlation between the “number of reversals” and any of the dependent variables. All the previous inconsistencies suggest that the TIN is not a good test to measure TP, and, therefore, no conclusions can be drawn in regards how well TP predicts student performance based on the results from the TIN. However, some conclusions about the particular PREDICTING ENGINEERING STUDENT’S PERFORMANCE 78 relationship between a particular sub-dimension of the TIN and student grades can be still interpreted with caution. There were four major findings than can be drawn from the analysis of the TIN. A first major finding was that two of the six sub-dimensions had no prediction of students’ grades; these are “student guessing” and “number of reversals.” From the other four sub-dimensions, the small prediction capacity of the sub-dimensions “number of items answered with an X” and “highest time-number” disappear when controlling for the sub-dimensions “number of corrects” and “number of items answered in the second half hour.” Consequently, these last two sub- dimensions were the only ones to have a unique predictive power of students’ grades. A second interesting finding is that the “number of items answered in the second half hour” correlates negatively with grades. This is contrary to what French (1948) predicted, hypothesizing that the amount of questions answered in the second half hour would be a measure of TP, meaning the more questions a student answers in the second half hour, the more persistent the student is. The interpretation of this sub-dimension was different in this study. The negative correlation with grades was interpreted as the possibility that students during the second half hour start feeling the pressure of time and start answering faster, guessing in some questions and making more mistakes. A third finding is that even the best sub-dimensions from the TIN in predicting student performance, which are the “number of corrects” and the “number of items answered in the second half hour,” are only good predictors for Calculus I, Linear Algebra and 1st Semester GPA. Controlling for other variables in the model, a one SD increase in the “number of corrects” is associated with an increase of 0.22 (p=0.012), 0.32 (p=0.002) and 0.15 (p=0.035) grade points, respectively. At the same time, one SD increase in the “number of items answered in the second PREDICTING ENGINEERING STUDENT’S PERFORMANCE 79 half hour” is associated with a decrease of 0.23 (p=0.010), 0.25 (p=0.013) and 0.16 (p=0.023) grade points, respectively. Finally, a fourth finding is that, when controlling for PSU scores, neither the “number of corrects” nor the “number of items answered in the second half hour” are significant predictors for any of the dependent variables, and therefore, TIN results are not providing any additional information to what is already provided by PSU scores. Discussion of Findings in Relation to Literature There is broad literature that supports the relationship between FDI, LI and TP and student performance. For FDI, it was seen that, without controlling for PSU scores, FDI might contribute to predictions of students’ grades for the more mathematics-intense courses. These results are aligned with the literature. Most of the reviewed studies in FDI found significant correlations between the GEFT and students’ grades in math-oriented courses (Bowlin, 1988; Tinajero & Paramo, 1997). In addition, the results of this study are similar to those of Witkin et al. (1977), who found that even though the GEFT correlates with students’ performance in mathematics and the sciences, this correlation is not significant when controlling for SAT scores (in our case, for PSU scores). Similar to FDI, LMI has been also associated with overall higher achievement and particularly higher mathematical achievement (Callaman, 2014; Cerda et al., 2011; Korkmaz, 2012; Nunes et al., 2007). While comparisons between the results of this study and the literature reviewed can be made for the cases of FDI and LI, it is harder to compare the results of this study with the existing literature for the case of TP, mainly because of the problems discussed in regards the TIN. Most of the studies analyzed during the literature review utilized questionnaires to assess a person’s ability to persist at a task. However, for this study, we needed to find a test that allows PREDICTING ENGINEERING STUDENT’S PERFORMANCE 80 assessing TP for selection purposes, and, therefore, self-reported questionnaires were not an option. The main comparison that can be made is between the results of the present study and French’s (1948) findings. As presented in Chapter Four, there are some similarities and some differences between the findings of the two studies. One similarity between the results of this study and that of French (1948) is that there was not a significant relationship between the variable “number of items answered with an X” and students’ performance for engineering students. On the contrary, there were important differences between the best sub-dimensions of the TIN in predicting student performance. French (1948) found the highest positive correlations between students’ grades and the sub-dimensions “highest time-number” and “number of reversals” from the TIN, while the results of this study suggest mostly negative correlations between “the highest time-number” and grades and no significant correlation between the “number of reversals” and any of the dependent variables. Implications for Practice Unfortunately, the implications for practice are minimal, given that none of the dimensions measured seem to contribute to predicting student performance beyond what the PSU already predicts. One motivation of this study was to identify additional measures that could predict student performance in a broader area of knowledge, such as communication skills and economics. However, FDI, LMI and TP measured through the GEFT-C, TILS and TIN may be good predictors for the more mathematics-intense courses, but grades can be better predicted by the PSU scores. Therefore, at least in the context of the T+I Program, there is not enough evidence to implement any of the previous three tests as additional measures for selecting students entering via this program. Nevertheless, there is also not enough evidence to totally discard these PREDICTING ENGINEERING STUDENT’S PERFORMANCE 81 measures and additional research is needed to be able to determine the relevance of FDI, LMI and TP in predicting student performance of engineering students. Propositions for Future Research There are several opportunities for future research, and most of them relate to improvements that can be made to the instruments utilized to measure FDI and TP. The TILS, on the other hand, worked fairly well in measuring LMI and, therefore there are not propositions in regards to additional research for this measure. First, the GEFT-C that was utilized to measure FDI presented important ceiling effects, which limited the possibility to discriminate between students in the more FI end of the spectrum. Even though the GEFT has been broadly validated for the general population, it did not work well for engineering students. Usually, students applying to engineering are already in the more FI end of the FDI spectrum as compared to the rest of the population, and the test did not allow for strong identification of the differences among these kinds of students. Maybe by knowing the differences among the students in the more FI end, the data will allow to better understand the real influence of FDI in determining student performance for engineering students. Second, regarding the TIN, which was utilized to measure TP, one recommendation is to improve the TIN test, and experts in this matter could better evaluate if this is possible or not. One possible improvement would be the development of a computerized version of the test which to facilitate the analysis of relevant information, such as tracking the time that students spend in the different questions, tracking of reversals and evaluating how much time the students spend in a question before marking it with a X. Another alternative to improve the existing TIN PREDICTING ENGINEERING STUDENT’S PERFORMANCE 82 is to find a different test to measure TP. Even though this option seems better and easier, for this study, it was not possible to identify a better alternative to the TIN. In addition to a better selection of the tests utilized to measure FDI and LI, a third recommendation for research is to follow up with students during the next years in their programs. It might be that following the students only for the first year is not enough to find the relationship between student performance and the results from the GEFT-C, TILS and the sub- dimensions from the TIN. Therefore, it may be worth following up with those students as they evolve in the engineering program. It is possible that, after several semesters, students’ grades start showing a stronger relationship with FDI, LI and TP. Finally, the last recommendation is to explore other measures that may be complementary to the PSU, such as critical thinking, grit or other constructs not related to mathematical ability, that have been correlated with student performance (Duckworth, Peterson, Matthews, & Kelly, 2007; Stupnisky, Renaud, Daniels, Haynes, & Perry, 2008). Final Conclusions Final conclusions need to take into consideration the problems regarding the tests selected to measure FDI and TP. For FDI, even though the GEFT-C had an important ceiling effect, it is still a very reliable measure, and, therefore, conclusions about the extent to which FDI can predict student performance can be made. However, for TP, the TIN showed important problems in terms of reliability and criterion validity, and, therefore, conclusions about the extent to which TP can predict students’ performance cannot be made. Taking the previous problems into consideration, the most important conclusions based on the results of this study are, first, that, even though FDI and LMI were good predictors for the more mathematics-intense classes, their capacity to predict student performance is much lower PREDICTING ENGINEERING STUDENT’S PERFORMANCE 83 than that of the PSU. Therefore, there is no additional information that these tests would provide to the selection process for students who enter the PUC School of Engineering via the T+I Program, and the implementation of such tests at PUC without serious revision would be unwarranted. Second, given the important validity and reliability problems with the TIN, the results from this study are not able to inform whether TP might predict student grades, and therefore, further research in regards better measures of task persistence is strongly recommended. Even more importantly, given the broad literature that suggests that TP (measured via self-reported questionnaires) is an important component of student performance, further research in relation to identify assessment measures of TP is needed for selection purposes. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 84 References Admisión y Registros Académicos. (n.d.). Retrieved from: http://admisionyregistros.uc.cl/ Andersson, H., & Bergman, L. R. (2011). The role of task persistence in young adolescence for successful educational and occupational attainment in middle adulthood. Developmental psychology, 47(4), 950. Baron, J., & Norman, M. F. (1992). SATs achievement tests, and high-school class rank as predictors of college performance. Educational and Psychological Measurement, 52(4), 1047-1055. doi:10.1177/0013164492052004029 Boe, E. E., May, H., & Boruch, R. F. (2002). Student Task Persistence in the Third International Mathematics and Science Study: A Major Source of Achievement Differences at the National, Classroom, and Student Levels. Philadelphia, PA: Center for Research and Evaluation in Social Policy, University of Pennsylvania. Bowlin, D. A. (1988). An investigation of the relationships between Field dependent/independent cognitive styles and sex, IQ, academic achievement, curriculum track selection and hemispheric preference in high school seniors. Dissertation Abstracts International, 49, 1405-A. Callaman, R. A. (2014). Logical Intelligence as Correlates of Academic Achievement in Advanced Algebra of Fourth Year High School Students. JPAIR Multidisciplinary Research, 17(1). Cerda, G., Ortega Ruiz, R., Pérez Wilson, C. E., Flores, C., & Melipillán, R. (2011). Inteligencia lógica y rendimiento académico en matemáticas: un estudio con estudiantes de Educación Básica y Secundaria de Chile. Anales de Psicología, Mayo-Sin mes, 389-398. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 85 Council of Education (CNED). (n.d.). Tendencia de Matrícula total por carrera genérica. Retrived from: http://www.cned.cl/public/Secciones/SeccionIndicesPostulantes/ CNED_IndicesTableau_MatriculaSistema.html?IdRegistro=005 DEMRE, Universidad de Chile. (2014). Vicerrectoría de Asuntos Académicos Departamento de Evaluación, Medición y Registro Educacional. Compendio Estadístico Proceso de Admisión Año Académico 2014. Duarte, J., Bos, M. S., Moreno, M., & Morduchowicz, A. (2013). Educational Equity in Chile: Trends 1999-2011. Inter-American Development Bank. Duckworth, A. L., Peterson, C., Matthews, M. D., & Kelly, D. R. (2007). Grit: Perseverance and passion for long-term goals. Journal of Personality and Social Psychology, 92(6), 1087- 1101. doi:10.1037/0022-3514.92.6.1087 Etchepare, G. C., Ortega, R., Pérez, C., Flores, C., & Melipillán, R. (2011). Logical intelligence levels and math performance: a study on primary and secondary Chilean students. Anales de Psicologia, 27(2), 389-398. French, B. F., Immekus, J. C., & Oakes, W. C. (2005). An examination of indicators of engineering students' success and persistence. Journal of Engineering Education, 94(4), 419-425. French, J. W. (1948). The validity of a persistence test. Psychometrika, 13(4), 271-277. Gardner, H., 1943, & ebrary, I. (1999;2000;). Intelligence reframed: Multiple intelligences for the 21st century. New York: Basic Books. Green, D. O., (2004, April). Affirmative action, conflict, and the University of Michigan: An insider’s perspective. Paper presented at the annual meetings of the American Educational Research Association, San Diego, CA. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 86 Gurin, P., Dey, E. L., Hurtado, S., & Gurin, G. (2002). Diversity and higher education: Theory and impact on educational outcomes. Harvard Educational Review, 72(3), 330-366. Harackiewicz, J. M., Barron, K. E., Tauer, J. M., & Elliot, A. J. (2002). Predicting success in college: A longitudinal study of achievement goals and ability measures as predictors of interest and performance from freshman year through graduation. Journal of Educational Psychology, 94(3), 562. Hastings, J. S., Neilson, C. A., & Zimmerman, S. D. (2013). Are some degrees worth more than others? Evidence from college admission cutoffs in Chile. ().National Bureau of Economic Research. doi:10.3386/w19241 Kepner, M. D. & Neimark, E. D. (1984). Test-retest reliability and differential patterns of score change on the Group Embedded Figures Test. Journal of Personality and Social Psychology, 46(6), 1405-1413. doi: 10.1037//0022-3514.46.6.1405 Koljatic, M., Silva, M., & Cofre, R. (2013). Achievement versus aptitude in college admissions: A cautionary note based on evidence from Chile. International Journal of Educational Development, 33(1), 106-115. Korkmaz, Ö. (2012). The impact of critical thinking and logico-mathematical intelligence on algorithmic design skills. Journal of Educational Computing Research, 46(2), 173-193. Martin, J. H., Montgomery, R. L., & Saphian, D. (2006). Personality, achievement test scores, and high school percentile as predictors of academic performance across four years of coursework. Journal of research in Personality, 40(4), 424-431. Nunes, T., Bryant, P., Barros, R., & Sylva, K. (2012). The relative importance of two different mathematical abilities to mathematical achievement. British Journal of Educational Psychology, 82(1), 136-156. PREDICTING ENGINEERING STUDENT’S PERFORMANCE 87 Nunes, T., Bryant, P., Evans, D., Bell, D., Gardner, S., Gardner, A., & Carraher, J. (2007). The contribution of logical reasoning to the learning of mathematics in primary school. British Journal of Developmental Psychology, 25(1), 147-166. Organisation for Economic Co-operation and Development. (2012). Equity and Quality in Education: Supporting Disadvantaged Students and Schools. Retrieved from http://dx.doi.org/10.1787/9789264130852-enhttp://dx.doi.org/10.1787/9789264130852- en Pallant, J. (2013). SPSS survival manual. London, England: McGraw-Hill Education (UK). Riquelme, G., Segure, T., & Yévenes, R. (1991, July-December). Version Experimental del Test de Inteligencia Lógica. Paideia, 16, 77-85. Sternberg, R. J., & Grigorenko, E. L. (1997). Are cognitive styles still in style? American Psychologist, 52(7), 700-712. doi:10.1037/0003-066X.52.7.700 Stupnisky, R. H., Renaud, R. D., Daniels, L. M., Haynes, T. L., & Perry, R. P. (2008). The interrelation of first-year college students' critical thinking disposition, perceived academic control, and academic achievement. Research in Higher Education, 49(6), 513- 530. doi:10.1007/s11162-008-9093-8 Tinajero, C., & Páramo, M. F. (1997). Field dependence-‐‑independence and academic achievement: A re-‐‑examination of their relationship. British Journal of Educational Psychology, 67(2), 199-212. Witkin, H. A. (1967). A cognitive-style approach to cross-cultural research. International Journal of Psychology, 2(4), 233-250. Witkin, H. A., Moore, C. A., Oltman, P. K., Goodenough, D. R., Friedman, F., Owen, D. R., & Raskin, E. (1977). Role of the field-dependent and field-independent cognitive styles in PREDICTING ENGINEERING STUDENT’S PERFORMANCE 88 academic evolution: A longitudinal study. Journal of Educational Psychology, 69(3), 197. Witkin, H. A., Oltman, P.K., Raskin, E., and Karp, S.A., 1971, & Demick, J. (2014). Group Embedded Figures Test Manual (2 nd ed.). Menlo Park, CA: Mind Garden, Inc. Zimmerman, A. P., Johnson, R. G., Hoover, T. S., Hilton, J. W., Heinemann, P. H., & Buckmaster, D. R. (2006). Comparison of personality types and learning styles of engineering students, agricultural systems management students, and faculty in an agricultural and biological engineering department. Transactions of the ASABE, 49(1), 311-317.
Abstract (if available)
Abstract
This study used statistical techniques to analyze field dependence-independence (FDI), logical-mathematical intelligence (LMI) and task persistence (TP) as possible predictors of engineering student’s performance in Chile. The purpose of this study was to identify additional measures to improve the selection criteria of students applying to the school of engineering at the Pontificia Universidad Católica de Chile via the Talent and Inclusion Program. The sample was composed of 65 students admitted via the Talent and Inclusion Program and 58 students admitted via regular admission, all of them entering engineering the 2015 school year. The 123 students were tested in FDI, LMI and TP using the GEFT-C (Group Embedded Figure Test–Computerized Version), the TILS (in Spanish, Test de Inteligencia Lógica Superior) and the TIN (in Spanish, Test de Ingenio Numérico), respectively. The data were analyzed utilizing different statistical techniques, first without controls and then controlling for PSU (in Spanish, Prueba de Selección Universitaria), the university’s current selection test. Findings from this study indicate that even though FDI and LMI can predict students’ performance in the most mathematics-intense courses and also GPA, their capacity to predict student performance is drastically smaller than PSU scores. In addition, and given some important problems of validity and reliability of the TIN, the results from this study were not conclusive as to whether TP might predict student grades. Given the broad literature that supports the relationship between TP and students’ performance (but uses self-reported questionnaires instead of assessments), it is strongly recommended that researchers pursue investigations to identify reliable and valid tests to assess TP that are not self-report and that can be used for selection purposes.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Assessing and addressing random and systematic measurement error in performance indicators of institutional effectiveness in the community college
PDF
Student engagement in an outperforming urban school as measured through cognitive, behavioral, and emotional indicators
PDF
Predictors of success in meeting the high school mathematics and English requirements for college entrance: a longitudinal study
PDF
A predictive valdity study: correlation of admission variables with program completion and student performance on the National Certification Examination in a physician assistant program
Asset Metadata
Creator
Stekel Schwarz, Yael
(author)
Core Title
Field dependence-independence, logical-mathematical intelligence, and task persistence as predictors of engineering student’s performance in Chile
School
Rossier School of Education
Degree
Doctor of Education
Degree Program
Education (Leadership)
Publication Date
09/28/2016
Defense Date
08/29/2016
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Engineering,field dependence-independence,logical-mathematical intelligence,OAI-PMH Harvest,student's performance,task persistence
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Hocevar, Dennis (
committee chair
), Tobey, Patricia E. (
committee member
), Torres-Torriti, Miguel (
committee member
)
Creator Email
stekelsc@usc.edu,yaelstekel@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-308063
Unique identifier
UC11280262
Identifier
etd-StekelSchw-4830.pdf (filename),usctheses-c40-308063 (legacy record id)
Legacy Identifier
etd-StekelSchw-4830.pdf
Dmrecord
308063
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Stekel Schwarz, Yael
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
field dependence-independence
logical-mathematical intelligence
student's performance
task persistence