A longitudinal analysis using auxiliary information to model retention in undergraduate students

Attrition is an issue for colleges and universities, and attempts to retain students are becoming more and more difficult. This study focuses on predicting student attrition of first time incoming (FTIC) students over a long time period. The population of this study consists of all FTIC students from Fall 2001. The students were followed 3.5 academic years to observe whether they experienced attrition. Exploratory data analysis was conducted to examine existing independent variables and some variables that were created to determine their contribution to the model. A discrete time hazard method was used to measure the timing of event occurrence. Cumulative GPA after one semester, number of major changes, major type, and minority status were selected to be included the model. Cross-validation was performed on Fall 2002 FTIC students to assess model fit. Overall, the model did a great job of predicting attrition of students over the long term.