An odds-ratio indifference-zone selection procedure for Bernoulli populations

We present a new sequential, eliminating procedure for selecting the best system in a single-factor Bernoulli-response experiment with an odds-ratio indifference zone, where best refers to the system with the largest probability of success on any given trial. Analytical results show that the proposed procedure is more efficient than existing procedures in that it requires fewer observations to identify the best system when only two systems are being considered. Empirical results show that the procedure is more efficient for comparing up to five systems when the specified odds-ratio indifference zone is greater than two.