|
Activity Number:
|
512
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #307045 |
|
Title:
|
Generalized Linear Mixed Models with Sparse Binary Outcome Data: Comparing Estimation Methods
|
|
Author(s):
|
Marie-Eve Beauchamp*+ and Robert W. Platt and James A. Hanley
|
|
Companies:
|
McGill University and McGill University and McGill University
|
|
Address:
|
1020 Pine Ave., W., Room 16-C, Montreal, PQ, H3A 1A2, Canada
|
|
Keywords:
|
generalized linear mixed models ; sparse data ; estimation methods ; penalized quasi-likelihood
|
|
Abstract:
|
The limits of estimation methods for generalized linear mixed models that are available in commercial software packages have not been studied systematically. The penalized quasi-likelihood (PQL) is among the most commonly used. However, PQL may produce severely biased variance component estimates for binary outcome data where the number of occurrences is low within each cluster. We compare via simulations the performance of PQL and other estimation methods, e.g. adaptive Gauss-Hermite quadrature, that are available in commercial packages and that presumably produce more accurate results than PQL. We focus on binary outcomes with a low number of occurrences. The bias in the parameter estimates is quantified in several settings. Data and model characteristics influencing the bias are also studied. Results to date will be presented.
|
