|
Activity Number:
|
251
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306119 |
|
Title:
|
Analysis of a Disease and Probability of Exposure Association Using a Replicated Error-Prone Exposure Assessment
|
|
Author(s):
|
Chengxing Lu*+ and Robert Lyles
|
|
Companies:
|
Emory University and Emory University
|
|
Address:
|
3448 N. Druid Hills Road, Decatur, GA, 30033,
|
|
Keywords:
|
beta-binomial ; bias ; logistic regression ; misclassification ; replicates
|
|
Abstract:
|
In environmental epidemiologic studies, it is common for a binary exposure to be assessed multiple times in a manner subject to misclassification. In a case-control setting, we focus on exploring the association between a disease and the probability of exposure given such replicates and in the absence of a gold standard for exposure. Assuming a beta distribution for the exposure probability, we obtain the estimated association by maximizing the marginal likelihood of the observed exposure replicates and the disease status. In simulation studies, we compare the performance of the proposed method with that of a logical but biased approach that replaces the unknown true exposure probability by the sample mean of the replicates. The proposed method is shown to be superior in terms of bias and confidence interval coverage. A real data example will also be presented for demonstration.
|
