|
Activity Number:
|
481
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Consulting
|
| Abstract - #307566 |
|
Title:
|
Caution When Using Covariate Adjustment in Mixed Effect ANOVA
|
|
Author(s):
|
Zhenxu Ma*+ and Paul Feder
|
|
Companies:
|
Battelle and Battelle
|
|
Address:
|
5005 King Ave., Columbus, OH, 43201-2693,
|
|
Keywords:
|
mixed effect model ; covariate adjustment
|
|
Abstract:
|
One of the reasons to include a covariate into an analysis of variance model is that the covariate may be responsible for some of the variation in the dependent variable. Therefore the use of covariate adjustment removes this variation from the error or random variance. The result is to improve sensitivity of the tests for treatment effects. Often the covariate is simply included into the analysis without any transformation. In this presentation, we show that if the slope of the relationship between the dependent variable and the covariate interacts with a random main effect, the variance for the error or random variance may increase with an extra variance being artificially added, and the results are then misleading. A proper way to center the covariate will effectively remove this extra variance.
|
