Effect of delayed observations on Bayesian decisions for binary data.

A Bayesian decision rule for early termination of an experimental study of binary responses is presented. This early termination occurs when the predictive probability of reversing the decision when utilizing the delayed observations is small. The proposed approach utilizes Bayesian inferential tools such as Bayes factors and predictive distributions. A simulation study is conducted to evaluate the performance of the proposed approach. Some guidelines are given to determine when the study should be terminated early and when the investigator should wait for delayed observations before making a conclusion.