Regressão de dados binários : distribuição Weibull

Abstract

In this work a new class of models for binary data based on Weibull distribution is introduced. A review is made of the most known linkage functions. This class of models has as special case the complementary log-log model and approximates well the logit and probit models. Three real data sets are given to compare the proposed model with many others. In one of these data sets the model is extended to allow multinomial data, that is, a discrete variable with more than two outcomes. The results are very good, because the estimation of parameters is quite simple and the model has shown to be very eficient.