Maxima of random samples from closed planar regions

Abstract

We will review the basic setup and structure of James-Stein estimate of the mean of a normal vector, and establish its robustness property under deviations from squared error loss as well as multinormality. Two applications will be discussed.