Asymptotic properties of the empirical BLUP and BLUE in mixed linear models. (English) Zbl 0901.62038
Summary: We show in a general mixed model that the best linear unbiased estimators (BLUE) of fixed effects, with unknown variance components substituted by the REML estimates, are jointly asymptotically normal with the REML estimates. We also prove that given sufficient information the empirical distributions of the best linear unbiased predictors (BLUP) of random effects, again with REML-estimated variance components, converge to the true distributions of the corresponding random effects. As a consequence, we obtain a consistent estimate of the asymptotic variance-covariance matrix of the REML estimates. The results require neither that the data is normally distributed nor that the model is hierarchical (nested).
MSC:
62F12 | Asymptotic properties of parametric estimators |
62J10 | Analysis of variance and covariance (ANOVA) |
62H12 | Estimation in multivariate analysis |
62E20 | Asymptotic distribution theory in statistics |