Abstract: This paper deals with the problem of estimating the probability of a correct selection (PCS) in location parameter models. Practical lower confidence bounds for the PCS in location parameter models are presented with a user's choice of dimension q(1<=q<=k-1) for computation, where k is the number of populations. It is shown that the larger the q, the better the lower bound, but the more complicated the computation. The result when q=1 coincides with Kim's (1986) result. A numerical example is presented to show that our lower bound with q=2 improves Kim's result considerably. With an appropriate modification, our result can be applied to location-scale parameter models with the scale parameter unknown.
Key words and phrases: Ranking and selection, probability of a correct selection, confidence region, location parameter.