A Two-Stage Procedure with Elimination for Partitioning Binomial Populations with Respect to a Control
by
Elena Buzaianu
University of North Florida
Coauthors: Pinyuen Chen (Syracuse University)

We propose a two-stage procedure for partitioning experimental binomial populations into two subsets according to their unknown probabilities of successes, with respect to a control binomial population. In stage I we use the subset selection approach studied by Gupta (1956) to partition and to eliminate extreme experimental populations. In stage II sampling is carried out from the control population and the experimental populations that were not partitioned at stage I yet. At stage II, following the indifference zone approach proposed by Bechhofer (1954), we partition the remaining experimental populations. The two-stage procedure has the advantage of eliminating “too inferior” or “too superior populations” at the first stage, and so reducing the average sample sizes from those extremely “bad” populations. The performance of the two-stage procedure is compared with the performance of the single-stage procedure based on the pure indifference zone approach.

Date received: April 12, 2007