Modelling two types of heterogeneity in the analysis of student success

Abstract

Student dropout is a worldwide problem, leading private and public universities in developed and underdeveloped countries to study the subject carefully or, as has recently been done, to analyse what drives student success. On this matter, different approaches are used to obtain useful information for decision-making. We propose a model that considers the enrolment date to the dropout or graduation date and also covariates to measure student success rates, to identify what the academic and non-academic factors are, and how they drive the student success. Our proposal assumes that there is one part of the population who is not at risk of dropping out, and that the part of the population at risk is heterogeneous, that is, we assume two types of heterogeneity. We highlight two advantages of our model: one is to identify the period of higher risk to dropout due to considering the academic survival time and the second is due to the inclusion of covariates that enable us to identify the characteristics linked to dropout. In this research, we also demonstrate the identifiability of the model and describe the estimation procedures. To exemplify the applicability of the approach, we use two real datasets.