|
Activity Number:
|
292
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #309959 |
|
Title:
|
A Latent Group Approach for Combining Matched and Unmatched Case-Control Studies
|
|
Author(s):
|
Mulugeta Gebregziabher*+ and Paulo Guimaraes and Wendy Cozen and David Conti
|
|
Companies:
|
MUSC and MUSC and University of California, Los Angeles and University of Southern California
|
|
Address:
|
135 Cannon St suite 303, Charleston, SC, 29425,
|
|
Keywords:
|
case-control ; conditional likelihood ; latent group ; polytomous conditional likelihood
|
|
Abstract:
|
In case-control studies, it is common practice to compare two or more different sets of controls with the same case group to validate or confirm a positive or negative finding. This usually involves fitting separate models for each case-control comparison, testing the homogeneity of the parameters and if appropriate obtaining a pooled estimate. But, fitting separate models tends to lead to a homogeneity test with inflated type-II-error rate and to a pooled estimate with larger standard error. The problem is compounded when one control group is matched and the other is unmatched. The alternative to fitting separate models is to use a multinomial model. However, available methods for combining matched and unmatched case-control data do not handle multinomial response. We propose a unified latent group approach that can be used for both binary and multinomial response case-control data.
|
