Abstract
Gupta has presented a method which, given samples of size n from each of k normal populations, selects a subset of the k populations which contains the population with the largest mean, with probability at least P* The method guarantees this P* no matter how small the differences between populations. The experimenter may only be interested in differences which are greater than some indifference. This difference allows either smaller subsets or a larger claimed P* for a given subset. An approximation using Gupta's and Sobel's tables and brief exact tables are presented.