A decision-theoretic approach to subset selection
Jan F. Bj⊘rnstad · Communication in Statistics- Theory and Methods · 1981
The multiple decision problem of selecting a random non-empty subset of populations, out of k populations, that are close in some sense to the best population is considered in a decision-theoretic framework. Uniformly optimal procedures for non-negative semi-additive loss are derived. A class of likelihood-ratio type of procedures is shown to be admissible for monotone additive loss.