Modelos de regressão para duas variáveis dependentes

Abstract

The use of multivariate analysis has great application in the more complex data analysis found recently in the literature. One of the factors can be attributed to the use of Bayesian analysis, which gives flexibility to the construction of new models. Many problems in which two possibly correlated variables were treated as independent, today can be solved without great computational costs using a bivariate distribution when the assumptions of this distribution are guaranteed. The objective of this work is to present four different bivariate models for multiple linear regression, based on the normal distribution, under the Bayesian approach, and to apply them to a set of real data. As a result, the proposed bivariate models adjusted better to the data analyzed when compared to the model that assumes independence of the response variables.