A Non-Symmetric Sequential Procedure for Selecting the Better of Two Binomial Populations.
Abstract
A sequential decision procedure is considered for the problem of selecting the better of two Bernoulli populations. The procedure uses a play-the-winner sampling rule based on the difference between the numbers of successes from the two populations. Exact expressions for the probability of correct selection and for the expected sample sizes are obtained. Tables of stopping constants which guarantee a certain probability requirement are computed, along with certain associated expected sample sizes. A method for choosing optimal stopping constants is proposed. The procedure is shown to compare favorably with some other play-the-winner rules. Along with several other play-the-winner procedures which have been proposed, it suffers from the deficiency that it has very large expected sample sizes when the probabilities of success of both populations are very small and close to each other. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1973
- Accession Number
- AD0766468
Entities
People
- Masanori Fushimi
Organizations
- Cornell University