Distribution of sum of squares and products matrices for the generalized multilinear matrix-T model

Abstract

The generalized multilinear model with the matrix-T error distribution is introduced in this paper. The sum of squares and products (SSP) matrix, as a counterpart of the Wishart matrix for the multinormal model, and the regression matrix for the errors and the observed as well as future responses are defined. The distributions of the regression matrix as well as the SSP matrix, and the prediction distribution of the future regression matrix and the future SSP matrix are derived. © 2002 Elsevier Science (USA).