A simulation study of some suboptimal bayesian sequential sampling schemes for choosing the better of two binomial populations

Abstract

Bayesian sequential schemes which use the posterior estimates of the unknown probabilities for choosing the better of two binomial populations are studied. Single sequential, group sequential and fixed sample size schemes are compared under criteria such as the probability of correct selection, expected sample size and expected number of observations on the poorer population which are obtained by Monte Carlo simulation. A modification of a rule due to Bechhofer and Kulkarni (1981) and the play the clear winner rule are also considered.

Keywords:

• Bayesian sequential design
• binomial selection
• expected sample size
• group sequential schemes
• Monte Carlo simulation
• probability of correct selection