Abstract:
|
Statistical procedures are developed for the simultaneous testing of multiple hypotheses in group sequential experiments such as group sequential clinical trials. As a result, a decision has to be made on each tested parameter instead of combining all tests into one hypothesis. Further, the maximum number of sampled groups is restricted. Under these conditions, we developed a new group sequential procedure based on the truncated sequential probability ratio test. This method resulted in an improved single-hypothesis testing that appeared to be more efficient than Pollock's method. Extending this sequential algorithm to multiple hypotheses by means of Holm-Bonferroni stepwise approach, we derive a truncated group sequential procedure for the simultaneous testing of multiple hypotheses that controls familywise Type I and Type II errors in the strong sense, and attempts to optimize the overall expected sample size, and it results in the reduced cost of medical treatment in clinical trials.
|