Modeling Students Performance using Cox and Parametric Accelerated Failure Time Models
Azme Khamis, Che Hamat, Mohd Asrul Affendi Abdullah · Scientific Research Journal · 2020
This study explored the use of survival analysis to investigate the Bachelor's degree students' performance based on GPA, entrance qualification, faculty, and course. The study considered the application of semi-parametric and parametric Accelerated Failure Time (AFT) models. The main objectives of the study are to identify the covariates that dominate students' performance via the Cox model, to investigate the performance of the Cox model based on the Proportional Hazard (PH) assumption, and to compare the performance of parametric AFT models using Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Corrected Akaike Information Criterion (AICc).Results revealed that the Cox model suggested the covariates of GPA, faculty, entrance qualification, and the course had affected student performance.PH assumption in the Cox model was violated.This analysis concluded that the Cox model provided a less accurate estimate of student performance and further study should be conducted with parametric AFT models.In parametric AFT models, the Log-normal AFT was chosen as the best model and can be used as an alternative model for estimating student performance at universities and other similar higher educational institutions.