Estimating the Precision of Time Period Effects in Longitudinal Models with Serially Correlated and Heterogeneous Errors

Abstract
A procedure is presented for obtaining unbiased estimators of the error dispersion matrix for mixed models with heterogeneous variances and serial covariances of various orders. While applicable to orthogonal as well as nonorthogonal data situations, the procedure is illustrated through an application to multi-treatment orthogonal data with serial correlation of the first order.