A sequential test for two binomial populations.

A sequential procedure is proposed for deciding which of two sequences of Bernoulli trials has the greater success probability. The procedure covers cases in which the two sequences are being observed at different rates, whether by design or otherwise, as well as the case of pairwise observation. The error probability of the procedure depends approximately only on the odds ratio of the two probabilities, a desirable property in deciding which of two medical treatments has the greater probability of success.