|
Activity Number:
|
25
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #304510 |
|
Title:
|
Retrieving Optimal Information from Parallel Data Using a Mean Coverage Controlled Procedure
|
|
Author(s):
|
Peter Hu*+ and Peggy Wong
|
|
Companies:
|
Merck & Co., Inc. and Merck & Co., Inc.
|
|
Address:
|
351 N. Sumneytown Pike, North Wales, PA, 19454,
|
|
Keywords:
|
minimizing mean confidence length ; deterministic optimization ; maximizing overall study power ; Meta-analysis
|
|
Abstract:
|
When pooling together individual point estimates and the corresponding confidence intervals from a set of biomarkers to gauge the overall study power, it's often of interest to shrink the variance estimates for these individual signals. We explored whether Uno et al's (2005, Biometrika) approach to constructing an optimal confidence region for a parameter of interest can be applied to constructing a set of confidence intervals for individual biomarkers with the shortest mean length. We show how the frequentist version of this method can be derived via deterministic optimization, but without loss of generalizibility or precision of the conclusion made by Uno et al under a joint Bayesian/frequentist paradigm. We also show how an optimal set of simultaneous confidence regions can be constructed using this idea.
(The authors thank Tianxi Cai and Len-Jie Wei for their helpful suggestions.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
