|
Activity Number:
|
78
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #304192 |
|
Title:
|
Extension of the Beta-Binomial Model for Cluster Randomized Trials When Adjusting for Baseline
|
|
Author(s):
|
Pamela A. Ohman Strickland*+
|
|
Companies:
|
University of Medicine and Dentistry of New Jersey
|
|
Address:
|
683 Hoes Ln W. Rm 218, Piscataway, NJ, 08854,
|
|
Keywords:
|
cluster randomized trial ; clustered binary data ; EM algorithm ; clinical trial
|
|
Abstract:
|
Analysis of results from randomized group trials presents many difficulties, particularly when the outcomes are binary in nature and these outcomes are measured pre- and post-intervention. A popular method for comparing (cluster) proportions between groups of clusters is through modeling of the proportions via a beta-binomial model, ignoring any baseline information. This work extends the beta-binomial model to allow for control of baseline proportions. Two approaches are taken to estimation of the effect of intervention: (1) an EM-algorithm embedded within a profile-likelihood approach and (2) approximation of the complete data likelihood embedded within an EM algorithm. These two approaches are implemented and compared with each other as well as with standard generalized linear modeling techniques.
|
