Some problems related to the empirical Bayers estimate proposed by Johns[1] (δn), and its modification (δn′) by Maritz[3], are considered in this artical. The main result concerns the comparison of the remainders R(G,δn) - R(G) and R(G,δn′)-R (G), where R(G,δn) and R(G,δn′) denote the "over-all" Bayes risk(under quadratic loss) of δn and δn′ respectively, and R(G) denotes the Bayes risk of Bayes estimate, and G is the prior distribution. |