Previsibilidade da evasão escolar num modelo de regressão beta inflacionado

Abstract

School dropout, a phenomenon present to a lesser or greater degree in the educational systems of several countries, accentuated in populations with greater economic or social vulnerability, and sometimes associated with them. Much is discussed about the phenomenon, little is noticed in the effective search for effective tools to combat it. The work finds that Brazilian school institutions, regardless of whether they are private or public, of the level of education, do not have consistent data control for effective analysis with technical rigor, therefore without predictability of these dropout rates, exhausting attention to recurrent predictors, such as grade and frequency without the culture of inclusion of behavioral variables of the individual, the institution and the geosocial context. In this search, a database of two Portuguese schools was obtained with consistent variables which are considered by specialists, pedagogues and in the educational literature, as explanatory of the phenomenon of school dropout. After performing descriptive statistics, graphs and appropriate tests, exponential logistic regression models were applied. Comparing models by tests; lrtest, Wald, correlation; observing the best fit characteristics, two models were obtained for modeling in the GAMLSS and betareg families.