Optimal construction of a selection of a subpopulation (Q1071416)

Consider a population whose members may be identified by realized values of a random variable (X,Y), distributed according to a probability distribution P(\(\theta)\), where \(\theta\) is an unknown parameter but the form of P(\(\theta)\) is known. Values of Y characterize members of the population more suitably than those of X, but they are to appear in the future. For example, in personal selection, each value of X represents an applicant's score in an admission test for an establishment and each value of Y is to appear as his achievement score after his entrance to the establishment. The problem of selecting a subpopulation of the given population, on the basis of observed values of X, is treated by the author in a decision theoretic setting. The existence of Bayes and minimax rules under the constraint defined by the selection criterion is proved, and a necessary and sufficient condition for a rule to be a Bayes solution is also obtained.