Test for independence of two multivariate regression equations with different design matrices (Q762856)

Many works were done in the literature on the problems of estimation and testing of hypotheses under the classical multivariate regression model. They were primarily done under the assumption that the design matrix remains the same for all variables. But this assumption is unrealistic in some occasions since the nature of some variables is different from others. The authors propose a locally best invariant (LBI) test for the independence of two multivariate regression equations with different design matrices. Asymptotic null distributions of the LBI test statistic and two other test statistics useful for testing the independence of two multivariate regression equations are derived. The asymptotic nonnull distributions of the test statistics associated with the above procedures under local alternatives are also derived.