A Multivariate Statistical Analysis of Major Change Patterns and Significant Factors that Influence Graduation Rates: A Case Study at California State University, Long Beach

Abstract

In 2015 the California State University system launched Graduation Initiative 2025 which aims to eliminate the equity gaps in degree completion and increase the average four-year graduation rate from 19% to 40% and the average six-year graduation rate from 57% to 70%. To support CSULB in meeting these goals, this study focuses on performing a multivariate statistical analysis to determine the effects of major change and various demographic and academic factors on timely graduation. The dataset was obtained from the Department of Institutional Research and Analytics and contained academic and demographic information on first-time freshmen accepted between 2009 and 2012. Due to high multicollinearity, dimensionality reduction was performed using Factor Analysis, and the data were analyzed using a combination of hypothesis testing, correlation analysis, multinomial logistic regression, and Fishers linear discriminant analysis.

This study found that students who changed majors graduated at a significantly higher rate than students who did not change majors. Students who changed majors at least once graduated at over double the rate of students who did not change majors. The multinomial logistic model suggested that a student’s academic performance, measured as a weighted linear combination of GPA, and the number of DFW, WU, and WE courses, is the most significant predictor of a student dropping out versus graduating within four or six years. STEM majors, Pell-eligible students, and local admission students are more likely to graduate as six-year graduates as opposed to four-year graduates. Fisher’s Linear Discriminant Analysis provided two liner discriminants that can be used to predict a student’s graduation status and uncover graduation trends within departments and colleges. The first linear discriminant can be used to separate students who exit without graduating from students who graduate, and the second linear discriminant can be used to separate four- and six-year graduates.