ADMISSIBLE BAYES PROCEDURES AND CLASSES OF EPSILON BAYES PROCEDURES FOR TESTING HYPOTHESES IN A MULTINOMIAL DISTRIBUTION.
Abstract
The purpose of the paper is to further study the problem of testing a simple hypothesis in a multinomial distribution. It is shown that for large k and near alternatives that for testing simple hypotheses one only need consider test procedures which have quadratic forms in N = (N sub 1,...,N sub k) as test statistics. A central limit theorem is proven for sums of quadratic forms of multinomials. Finally it is shown that the chi-square test, likelihood ratio test and all quadratic forms are Bayes and admissible for fixed k and n.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 11, 1966
- Accession Number
- AD0642636
Entities
People
- Carl Morris
Organizations
- Stanford University