Accession Number : AD0675641

Title :   A CLASS OF SEQUENTIAL MULTIPLE DECISION PROCEDURES,

Corporate Author : PURDUE UNIV LAFAYETTE IND DEPT OF STATISTICS

Personal Author(s) : Barron,Austin M.

Report Date : SEP 1968

Pagination or Media Count : 111

Abstract : Consider k populations Pi sub 1, Pi sub 2,..., Pi sub k where each Pi sub i has an observable random variable which depends on some parameter theta sub i. The problem then is to define sequential multiple decision procedures, which select a subset Pi sub 1, Pi sub 2,..., Pi sub k such that the population with the largest (or smallest) mean is included with a prescribed probability P*. Two types of procedures are considered. The first is a non-eliminating type which takes observations from each population at each stage until a decision (to select or reject) has been made about all the populations. The second, an eliminating type, stops sampling from a population when a decision has been reached about that population. The first two chapters deal with normal populations when the parameters in question are the means. The last chapter offers some generalizations of the procedure and some related problems. (Author)

Descriptors :   (*STATISTICAL ANALYSIS, DECISION THEORY), (*DECISION THEORY, SEQUENTIAL ANALYSIS), STATISTICAL DISTRIBUTIONS, POPULATION(MATHEMATICS), NUMERICAL METHODS AND PROCEDURES, MONTE CARLO METHOD, MINIMAX TECHNIQUE, PROBABILITY, SAMPLING

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE