|
Activity Number:
|
162
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #304420 |
|
Title:
|
Optimal Two-Stage Design When Adapting Between k Sample Sizes
|
|
Author(s):
|
Hong Wan*+ and Susan Ellenberg and Keaven Anderson
|
|
Companies:
|
Merck & Co., Inc. and University of Pennsylvania and Merck & Co., Inc.
|
|
Address:
|
P.O. Box 1000, North Wales, PA, 19454,
|
|
Keywords:
|
sample size re-estimation ; adaptive design ; optimal design
|
|
Abstract:
|
Several adaptive design methods have been proposed to re-estimate sample size using the observed treatment effect at the first stage. The algorithms used in some methods can be inverted to reveal the treatment effect at the interim analysis. We propose using a step function of observed treatment difference for sample size re-estimation to lessen the information on treatment effect revealed. This method applies calculation methods used for group sequential design. We minimize expected sample size among these designs and compare efficiency to the fully optimized two-stage design proposed by Lokhnygina and Tsiatis (2008). The tradeoff between efficiency versus the reduction in information revealed will be discussed.
|
