Comparação de testes sobre estruturas de covariância de populações normais

Abstract

In some studies there is interest in testing the structure of the variance and covariance matrix, as in the context of several multivariate or modeling techniques. In light of this, the importance of using hypothesis tests on covariance structures is emphasized and a detailed and comparative study of power and error type I of some existing tests is made. Considering the scenarios with different numbers of variables (p = 2;4;8;16;32;64) and sample sizes (n = 5;10;20;30; 50; 100). In order to evaluate the identity of the proposal of Ledoit e Wolf (2002) was the most indicated, for the sphericity the version of John (1972) modified by Ledoit e Wolf (2002) followed by the proposal of Box (1949) were the best performance. In the study for any structure the most recommended test is a version also proposed by Ledoit e Wolf (2002).