Abstract
This paper considers the problems of selecting the r best of k events (1≦r<k) in a multinomial distribution. A statistical procedure for selecting the r best events in an unordered manner is proposed from the indifference zone point of view. The experimenter specified two constants δ^* and P^* where 1/(^k_r)<P^*<1. The smallest sample size is determined so that the probability of selecting the r best events is at least as large as P^* whenever the r-th largest cell-probability is at a distance not less than δ^* from the (r+1)-th largest cell-probability. Graphs are given for k=3,4,5 to assist the experimenter in designing the experiment.
