Title: On estimation and prediction in multivariate mixed linear models
Authors: Tatjana von Rosen - Stockholm University (Sweden) [presenting]
Abstract: Mixed linear models have been extensively studied due to their intensive use in practical data analysis when the response is univariate. Much attention has been paid on the equality of fixed parameter estimates and the equality of best linear unbiased predictors of random effects under two mixed models with different covariance matrices. The latter can be an important question, for example, in small-area estimation. Nevertheless, these issues have not been considered in the context of multivariate mixed linear models. Therefore, the goal is to extend the results on the equality of best linear unbiased estimators and best linear unbiased predictors to the multivariate situation. For simlicity, the balanced multivariate mixed linear models will be studied.