Dados do Trabalho

Título

STUDY OF UNIVERSITY DROPOUT USING COMPETING RISKS MODEL

Resumo

High education systems face dropouts and late graduations and then have to deal with economic and social issues. Our goal is to study students dropout from the Bachelor Applied Mathematics course at University of Sao Paulo, whose dropout rate is increasing and worrisome, and also the covariates associated to the risk of dropout and to the risk of graduation. In our proposal we assume that the (latent) failure time is a continuous variable following the proportional hazard model that can just be observed in discrete times (semesters). Since dropout and graduation are competitive events we assume a competing risk survival modeling. We use the log complementary transformation on the risk function to obtain a linear function of the parameters implying that we can fit a generalized linear model with binomial error structure to estimate the regression coefficients. The Bayesian approach is considered to deal with covariate selection and estimation of the parameters. We conclude that the risk of graduation is higher for those who was admitted by first call, that is the risk of dropout is higher for those who was admitted after the first call. Despite this conclusion, in general the covariates may not explain the events for this course in particular.

Palavras-chave

Bayesian model averaging, Competing risk, University dropout