Sample Size Requirements to Control Bayesian Error Probabilities in Superiority Trials

In this paper, we consider the problem of selecting the opti- mal sample size by controlling the risk of errors in statistical hypothesis testing. Bayesian versions of the frequentist Type I and Type II error probabilities can be obtained by exploiting the two-priors approach. The criterion we consider is based on the control of a weighted average of both errors, that allows to modulate the contribution of each error in the de- termination of the sample size. The procedure is described by considering a two-arm superiority trial based on binary data