Robustness of Bayes decisions for normal and lognormal distributions under hierarchical priors

In this paper we derive the Bayes estimates of the location parameter of normal and lognormal distribution under the hierarchical priors for the vector parameter, . The ML-II ε-contaminated class of priors are employed at the second stage of hierarchical priors to examine the robustness of Bayes estimates with respect to possible misspecification at the second stage. The simulation studies for both normal and lognormal distributions confirm Berger’s (1985) assertion that form of the second stage prior does not affect the Bayes decisions.