|
Activity Number:
|
221
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #304769 |
|
Title:
|
Design Effects with Generalized Linear Mixed Effects Model
|
|
Author(s):
|
Qiaohao Zhu*+ and Keumhee C. Carriere
|
|
Companies:
|
University of Alberta and University of Alberta
|
|
Address:
|
179 MacEwan Road, Edmonton, AB, T6W1K3, Canada
|
|
Keywords:
|
Design Effect ; Generalized Linear Model ; Generalized Linear Mixed Model ; Intra-cluster correlation coefficient ; Goodness-of-Fit Test
|
|
Abstract:
|
Design effect was first introduced by Kish (Kish, 1965) as a measure of efficiency when comparing complex sampling survey designs to simple random sampling design, and was also used to measure the impact on model-based inference with complex survey designs. We will apply design effect as a measure and adjustment factor of comparing Generalized Linear Mixed-Effects Model (GLMM) to the corresponding model without random effects (GLM). We propose a procedure of estimating intra-cluster correlation coefficient (ICC) from the GLMM through data linearization method, and estimate the design effect based on the estimated ICC. As an application of the GLMM design effect, we propose a goodness-of-fit test for logistic regression model with random effects by modifying the Hosmer-Lemeshow goodness-of-fit test incorporating the estimated design effect.
|
