Análise de sobrevivência aplicada à evasão de alunos de graduação

Abstract

Evasion in undergraduate courses is a reality in most Higher Education Institutions, both public and private, with their alarmingly high rates. In this sense, the present work to study the time until the evasion of undergraduate students at the Federal University of Uberlândia (UFU) through the techniques of survival analysis. The data set represents the situation of students in the courses located on the Santa Mônica and Umuarama campuses, in the city of Uberlândia. The covariates set consisted area of knowledge the course, from of entry in course, place of birth, sex, age of students and the academic performance coe cient of student, named CRA (coe ciente de rendimento acadêmico). Firstly, the survival curve was estimated using the Kaplan-Meier estimator and the curves of each covariate were compared using the logrank test. Cox's model was tted for both campuses. Some covariates presented violation of the proportional hazards assumption, according to the test, but with graphical analysis the assumption was accepted. Alternatively, Aalen's additive model and semiparametric additive risk model were tted, allowing time-dependent covariates and also time-independent. In general, the results were consistent for both campuses, for example, the dropout rate for SISU students is higher when compared to the entrance exam, the male showed a higher dropout rate, increasing age causes an increase in the dropout rate and the rate of permanent evasion with the increase in CRA. The place of birth covariate was the only one that presented a divergent result, on the Umuarama campus the dropout rate is lower for students from Uberlândia, while on the Santa Monica campus the rate is higher.