Emory University, Manager, Decision Support & Data Management
Traditional models utilizing predictive analytics have focused on undergraduate populations of IPEDS defined first-time, full-time freshman yield probabilities, retention, and graduation rates. Yet as is well documented in the literature, this only encompasses a small population of students in higher education. For graduate education, these types of models are rarely employed since they frequently lack centralized admissions offices and are managed departmentally. However, empirical evidence demonstrates that the same analytical procedures that are used in undergraduate education can be applied at the graduate and professional level to achieve similar results. In many cases, such as in legal, accounting, and medical education, the measure of student success is a high stakes measure of passing a professional board.
This presentation focuses on three ways of comparing student success analytical models across levels, demonstrating a more universal approach to student learning. These three ways of comparing and contrasting student success models illustrate similarities in estimating and forecasting student success. Furthermore, utilizing these models can help to identify students who may be at-risk, provide intervention strategies, and evaluate these strategies. Incorporating predictive analytics and program evaluation can help to ensure high-impact, low-cost pedagogical approaches are employed to assist students with their educational journeys.
Justin Shepherd has a Ph.D. in higher education policy from Vanderbilt University and is credited with dozens of conference presentations (AIR, SAIR, ASHE, AERA) over the past five years. His work blends econometric tools with information technology to produce statistical estimates for program evaluation, behavioral modeling, and predictive analytics. His background focuses on quantitative research methods, survey research, and hypothesis testing. From this foundation, he integrates concepts from business intelligence, data visualization, and data sciences to turn data into actionable information.