Abstract:
|
The Closure Principle of Marcus, Peritz, and Gabriel (1976) is re-visited, and a new implementation of this principle is devised which is shown to provide improvement of power over other closed testing procedures in many common settings. Such settings arise in medical studies when the hypotheses being tested correspond to treatment effects that are expected to trend similarly. As an example, a drug that is expected to quicken the time to improvement of a disease is usually also expected to prolong the improvement time. In this paper, we first discuss the implementation of the proposed procedure in the case of two hypotheses where the test statistics are independent. We prove that the procedure maintains strong control of the familywise error rate (FWER) under mild regulatory assumptions on the distribution of the test statistics. We then devise an extension of Hochberg's step-up procedure using the same principle. We compare the extended procedure to the original Hochberg procedure under various non-null configurations. We also show how the procedure can be extended to correlated test statistics while maintaining strong control of the FWER. Finally, we propose an extension for more than two hypotheses.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.