Modelo de regressão beta retangular para análise de dados com medidas repetidas

Abstract

The independence between observations is a usual hypothesis for the application of numerous statistical techniques and is generally adequate when only one value is observed for each sample unit. However, it is common to have experiments with repeated measures, that is, when there is more than one observation for each of these units. Considering the structure of repeated measures, the possible existence of a correlation between the observations of the same unit and cases in which the response variable is restricted to the interval (0,1), we propose the modeling of the mean of a rectangular beta regression model from the Generalized Estimation Equations under the assumption of homogeneity of the precision parameter. Under this model we construct diagnostic techniques such as leverage points, Cook’s distance and local influence, and present an example to illustrate the methodology developed.