PhD. Dissertation Defense Announcement: Jay Stefanelli - Using Binary Logistic Regression to Analyze the Predictive Validity of Undergraduate GPA and Standardized Test Scores in Forecasting Master’s Student Degree Completion

Thursday, January 14, 2021 2:00pm - 3:00pm

Zoom Meeting - Virtual

Wedged between undergraduate and doctoral students exists the oft-understudied population of master’s students. The dearth in master’s-specific research is astounding considering master’s degrees are the fastest growing degree credential. Why is there such a lack of information about master’s students from a national perspective in comparison to other degree-seeking students? Perhaps it is because assessment, like IPEDS, and prestige, like US News rankings, are not tied to it. In simple terms, there isn’t a carrot to incentivize it or a stick to mandate it. The absence of research has far-reaching implications and manifests primarily as a disconnect between the efficacy of admissions and anticipated student degree completion, but has further implications on our grasp of master's student graduation rates and time to degree. This study bridges this gap by employing binary logistic regression to analyze the relationship between the independent variables; admissions-, demographic-, and environment-based attributes, and the dependent variable master’s degree completion. In looking specifically at the traditional measures of success within the master’s admissions process; undergraduate GPA and standardized test scores, statistical significance was not found. Within the demographic attributes; age, race/ethnicity, gender, citizenship, and residency, only age was statistically significant. Simply, the odds of graduation decrease as students grow older. Finally, within the environment attributes; institution, academic discipline, attendance pattern, start term, representation by race/ethnicity, and representation by gender, the model indicated statistical significance across categories within institution, academic discipline, attendance pattern, and start term. Certain institutions and academic disciplines are associated with greater odds of graduation than others, full-time students more likely to graduate than their part-time peers, as are students that start their program in the Fall term when compared to those that start in the Spring. This is the first-ever, multi-university study to employ a statistical model to analyze the predictive validity of the traditional measures of success, and accompanying demographic and environment attributes, in forecasting degree completion. It is also the most comprehensive in terms of program breadth. Additionally, this study provides university administrators with a comparative benchmark for both master's student graduation rates and time to degree.


Committee: Jerome Kukor, Ph.D., Ralph Gigliotti, Ph.D., Hironao Okahana, Ph.D., Karen Stubaus, Ph.D.



Who to contact:

Matt Winkler