Research Tools for Analysis of Dichotomous Outcomes in Education: A Comparison
Abstract
This study aimed to provide a guideline for the proper use of a survival model in an educational research field in which discrete and continuous models often need clarification. To achieve the goal, we compared discrete- (logit) and continuous-time (complementary log-log) survival models by considering the hazard rate. We simulated data for various combinations of time metrics, censoring proportions, and sample size. The study results recommend discrete models for cases with large time metrics, small proportions of censored observations, and small samples. The study highlights the importance of adopting a proper model in using survival analyses.