|
Activity Number:
|
457
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 7, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
ENAR
|
| Abstract - #300315 |
|
Title:
|
Power for Clustered Gaussian Data
|
|
Author(s):
|
Jacqueline L. Johnson*+ and Keith E. Muller and Diane J. Catellier and Mildred M. Maldonado-Molina and Kelli A. Komro
|
|
Companies:
|
Novartis Pharmaceuticals and University of Florida and The University of North Carolina at Chapel Hill and University of Florida and University of Florida
|
|
Address:
|
One Health Plaza, East Hanover, NJ, 07936,
|
|
Keywords:
|
group randomized trials
|
|
Abstract:
|
Even though the data analysis in most cluster studies with Gaussian data uses some type of weighting to account for imbalance, current methods for computing power assume equal cluster sizes, for which exact power can be computed. Unbalanced data typical of cluster designs can make actual power differ substantially from the planned power. We describe how to align power calculations with a hybrid method of data analysis which combines mixed model estimation of intraclass correlation with exact transformation to a univariate model for approximate hypothesis testing. Analytic results, enumerations, and simulations all support the approach. The method allows accounting for within-cluster covariates. A power analysis for a study of adolescent drinking behavior illustrates how easily the method can be implemented with standard data analysis and power software.
|
- 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 2008 program |
