Tatjana von Rosen: Estimation in Models with a Kronecker Product Covariance Structure

We consider a (p×q)-dimensional random matrix X distributed normally with
mean my and covariance matrix Sigma= Psi (X) Fi, where Psi: q× q, Fi :p × p
and (X) is the Kronecker product
are assumed to be positive definite but unknown. Based on a sample of matrices under different structures on the parameter matrices maximum likelihood estimators are obtained via flip-flop algorithms. In particular, the convergence of the algorithm to a unique solution is discussed.

(In collaboration with M. Srivastava and D. von Rosen)