Some Efficient Simple Rules in Gamma-minimax Estimation

Brani Vidakovic, David Rios Insua
Duke University, Universidad Politencica de Madrid

Nov 30 1992

We discuss the T-minimax estimation of normal means under squared-error loss. Since this problem is computationally intensive, we study linear T-minimax estimation. we identify contexts in which linear rules are adequately efficient, in terms of the ratio of their risk to the optimal risk. We also provide strategies that may work in other contexts.


Normal means, robust Bayesian analysis, T-minimax, linear rule, Fisher information, almost T-minimax


PDF icon 1993-10.pdf