Bayes-P* Subset Selection Procedures for the Best Population.
Abstract
Two new selection procedures, called nonrandomized and randomized Bayes-P* procedures are defined for selecting a small nonempty subset of k populations which contains the best population. It is shown that these procedures have some optimal properties. If we restrict attention to the class D(D*) of all nonrandomized (randomized) selection procedures, which satisfy the PP*-condition, that is the posterior probability of a correct selection, for any given observation X = x, is not less than P*, a predetermined number between l/k and l, then these two new selection procedures are shown also to be Bayes decision procedures in the class D and D* respectively, provided that some regularity conditions are satisfied. Robustness of these procedures and comparisons with some other selection procedures are studied by using Monte Carlo simulations. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1984
- Accession Number
- ADA140179
Entities
People
- H. M. Yang
- Sumedha Gupta
Organizations
- Purdue University