Predicting Student Success: A Logistic Regression Analysis of Data From Multiple SIU-C Courses

The objective of this report is to improve prediction techniques regarding the future performance of students in select university courses through the utilization of multiple logistic regressions. This is achieved with the aid of statistical computing software which applies forward step-wise variable selection methods that identify influential variables sufficient to accurately predict student success. Once a logit model is constructed with the required parameters and predictors, the inverse logit function outputs a probability of student success. In all cases, logistic prediction models matched or exceeded the performance of current prediction methods while using an equal or lesser number of explanatory variables. These findings show that current prediction methods can improve by using a statistically justified procedure. It also suggests the inefficacy of some predictors used to currently estimate student performance