Testes de bondade de ajuste para distribuição Birnbaum-Saunders baseados na informação de Kullback-Leibler.

Abstract

In this work, we propose goodness-of-fit tests based on the Kullback-Leibler information for the Birnbaum-Saunders model, in which we initially considered uncensored data and, in a second moment, we considered data with progressive censorship of type II. For each case, the respective test statístics were obtained, as well, as the process of estimation of the parameters of the model was discussed. In the uncensored case, the sizes and the power of the test were obtained for diferent sample sizes and different alternatives for the risk function by means of Monte Carlo simulations and, in addition, we compared such powers with those of the classic tests of goodness of fit, namely the Cramér-von Mises test and the Anderson-Darling test. In the case of progressive type II censorship, the sizes and powers of the test were also evaluated throught simulation studies considering different sample sizes, different sample sizes, different censorship schemes and different alternatives for the risk function. Finally, we made applications with real data sets.