|
Activity Number:
|
364
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section*
|
| Abstract - #301390 |
|
Title:
|
A Weighted Least Squares Model for the Analysis of Clustered Binary Data
|
|
Author(s):
|
Mitchell Rosen*+
|
|
Affiliation(s):
|
ICON Clinical Research
|
|
Address:
|
212 Church Rd., North Wales, Pennsylvania, 19454, USA
|
|
Keywords:
|
Weighted least squares ; Clustered binary data ; Correlated binary data ; Beta-binomial distribution ; Intracluster correlation
|
|
Abstract:
|
The weighted least squares (WLS) method is proposed for estimating the per unit agreement rate in clustered binary data. Two WLS estimators, which are algebraically equivalent to the ratio estimator and the weighted estimator, are recommended. A key advantage of WLS estimators is that confidence intervals are based on students' t distribution rather than on the normal approximation. The results of a computer simulation suggested that the ratio estimator (with WLS standard error) is appropriate when the intracluster correlation is less than 0.2, while the weighted estimator is optimal when the intracluster correlation is 0.2 or greater. The simulation results also indicated that asymptotic sample size and power calculation methods have low accuracy when the success rate is outside the range 0.3< =p< =0.7.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |
