Modelo de regressão beta aplicado à incidência da Covid-19 no Brasil

Abstract

The beta regression model is often used to model data that has support defined on the interval (0,1). This model gained significant attention since its presentation by Ferrari e Cribari-Neto (2004). Maximum likelihood (ML) method is generally used to estimate the parameters of the beta regression model, however, when the sample size is small, the biases of the estimators cannot be neglected. Therefore, it is necessary to correct these biases. Among the existing corrections for ML estimators, the bootstrap correction can be cited. In this text, simulation study was conducted to verify the performance of ML estimators and bootstrap-corrected ML estimators of the parameters of the beta regression model. Additionally, the beta regression model was applied to COVID-19 data from the 26 brazilians states, including the Federal District, in 2020.