Desempenho de alguns Estimadores Ridge no Modelo de Regressão Birnbaum-Saunders.

Abstract

The Birnbaum-Saunders regression model has been applied in di erent areas of knowledge, for example, in the analysis of competence and reliability data. However, the presence of multicollinearity in the model negatively a ects a variance of the maximum likelihood estimator. In the literature, we nd several methods to minimize such e ects, one of which is the ridge estimator it provides, which has been extensively studied and is e cient in the presence of multicollinearity. In this work, we propose ridge estimators for the regressive parameters of the Birnbaum-Saunders model. We evaluated the proposed estimator in terms of the mean square error and showed that it has a lower mean square error than the maximum likelihood estimator. We also present some evaluation proposals for the shrinkage parameter. We evaluated the proposed estimators through Monte Carlo simulation.