Let X be a vector of independent components with mean vector θ. We assume that the distribution of the jth component is of the form [formula omitted], i.e. a variant mixture of normal distribution. We show that certain explicit James-Stein type estimators are minimax for the problem of estimating the vector [formula omitted] under the loss function [formula omitted].
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- James-Stein estimation
- Location parameters
- decision theory