On the Allocation of Treatments in Sequential Medical Trials

Summary Several allocation rules are compared by assessing their effectiveness in discovering which of two treatments has the higher probability of success. The number of observations needed to obtain a given level of precision can be minimized by using a fixed-sample rule, but this involves too many applications of the inferior treatment. It is demonstrated that sequential allocation rules can achieve a similar pattern of error probabilities for a small fraction of the expected cost to the volunteers employed in the experiment.