Simultaneous estimation of parameters under entropy loss (Q1086927)

The authors consider the simultaneous estimation of the mean-vector, and the vector of reciprocals of the means from p independent gamma distributions. Under the entropy loss, the admissibility of the best scale invariant estimator is proved when \(p=2\) and some conditions on the shape parameters hold. When \(p\geq 3\), its inadmissibility is proved and several improved estimators are obtained. The authors also obtain a result showing the inadmissibility of the generalized Bayes estimators of a vector consisting of certain powers of the scale parameter. An analogous loss to the above was discussed in \textit{W. James} and \textit{C. Stein} [Estimation with quadratic loss. Proc. 4th Berkeley Sympos. math. statist. Probab. 1, 361-379 (1961)], for the estimation of the variance-covariance matrix of a multinormal distribution.