An optimal sequential selection procedure

Abstract

The problem is to select on the basis of repeated observations from k populations “the best” one guaranteeing the indifference zone requirement on the probability of correct selection. We propose a sequential selection procedure which asymptotically minimizes the expected sample size for all parameter values, including the indifference zone, when probability of correct selection is close to one. It is shown to be substantially superior in asymptotic sample size behavior to existing procedures.

Keywords:

• Selection procedure
• indifference zone
• adaptive likelihood ratio
• asymptotic optimality