In this paper, we are interested in the problem of simultaneous inference and selection from among populations in comparison with a standard or control. The populations are denoted by . The random variable associated with is assumed to have the density .

A nonparametric empirical Bayes approach is used to construct the selection procedure based on data from past n stages and the present stage. It is shown that this empirical Bayes procedure is asymptotically optimal with a rate of order . The results are applicable to data arising from life-test experiments.