Accession Number : AD0626603

Title :   SUFFICIENCY OF THE TERMINAL DECISION IN A SPRT,

Corporate Author : RAND CORP SANTA MONICA CALIF

Personal Author(s) : Bussgang,J. J. ; Marcus,M. B.

Report Date : JAN 1966

Pagination or Media Count : 7

Abstract : Consider a situation in which an experimenter observes a stochastic process starting at time t = 0. A continuous sequential test of simple hypotheses concerning the underlying distribution of the process is performed and terminates at some time t = or > zero. At termination, a choice is made between the alternate hypotheses in accordance with the test procedure. The test procedure is designed to produce this choice within specified probabilities of error. The experimenter may, however, be concerned whether he can also utilize his knowledge of the test duration of the particular test he performed in order to increase his knowledge about the underlying distribution of the process. The results of this paper demonstrate that for a sequential probability ratio test (SPRT) he can not. In fact the terminal decision and the test duration provide no more information than the terminal decision alone.

Descriptors :   (*SEQUENTIAL ANALYSIS, DECISION THEORY), STOCHASTIC PROCESSES, STATISTICAL TESTS, PROBABILITY

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE