|
Activity Number:
|
548
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #305443 |
|
Title:
|
A New Gatekeeping Strategy for Hierarchical-Structured Hypotheses in Clinical Trials
|
|
Author(s):
|
Junyuan Wang*+ and Yan Zheng and Guanghan Liu
|
|
Companies:
|
The Medicines Company and University of Minnesota and Merck Research Laboratories
|
|
Address:
|
8 Campus Drive, Parsippany, PA, 07054,
|
|
Keywords:
|
multiplicity adjustment ; weighted Simes ; Hochberg method
|
|
Abstract:
|
Gatekeeping strategies have been proposed to exploit the hierarchical structure of primary and secondary hypotheses when there are multiple primary hypotheses. Dmitrienko et al. (2003) suggested a weighted Simes method that provides high power compared to other gatekeeping methods. However, it satisfies neither of the two properties suggested in their paper. We propose two two-stage gatekeeping strategies that are simpler, with comparable power for the primary hypotheses and higher power for the secondary hypotheses. The proposed two-stage methods satisfy (1) overall significance of the primary family will not be affected by the significance of the secondary hypotheses and (2) only if at least one of the primary hypotheses is significant, can one proceed to test the secondary family. The properties of the new gatekeeping methods are demonstrated through simulation.
|
