Asymptotic optimality of hierarchical Bayes estimators and predictors (Q1193955)

The present paper deals with hierarchical Bayes (HB) estimation of a vector of means \(\theta\) in the \(k\)-dimensional normal linear model with known or unknown first stage variance component. Certain HB estimators of \(\theta\) are proved to be asymptotically optimal in the sense that the difference in the Bayes risks of the estimator and the `true' Bayes estimator converges to zero as \(k\to \infty\). The results are then extended to prove the asymptotic optimality of certain HB predictors. As an application, certain HB predictors of \textit{T. W. F. Stroud} [Bayes and empirical Bayes approaches to small area estimation. In: R. Platek, J. N. K. Rao, C. E. Särndal and M. P. Singh (eds.), Small Area Statistics, 129-137 (1987)] and \textit{M. Ghosh} and \textit{P. Lahiri} [A hierarchical Bayes approach to small area estimation with auxiliary information. Joint Indo-U.S. Workshop Bayes. Anal. Statist., Econometrics. Bangalore/India (1988)] arising in the context of small area estimation are shown to possess the asymptotic optimality property.