Multivariate Model with Correlated Observation Units

Abstract

In this paper we consider $p$ characteristics at $n$ time-points. These observations can be represented as $n$ random $p$-vectors $X_1, \ldots, X_n$. We assume that a $pn$-vector ${\rm vec}({\bf X}) = \left(X_1^T, \ldots, X_n^T\right)^T$ is distributed normally and follows a Kronecker product covariance structure. New methods to test linear hypotheses in the considered model are presented.