Online Program Home
My Program

Abstract Details

Activity Number: 490
Type: Contributed
Date/Time: Wednesday, August 3, 2016 : 8:30 AM to 10:20 AM
Sponsor: Biopharmaceutical Section
Abstract #320057 View Presentation
Title: Small Sample Inference in Imbalanced Cluster Randomized Clinical Trials with Binary Outcomes
Author(s): Dong Hyun Ahn* and Judith D. Goldberg
Companies: New York University and New York University School of Medicine
Keywords: Cluster Randomized Trials ; Clustered Data ; Generalized Estimating Equations ; Cluster Size Imbalance ; Coefficient of Variation ; Bias-Corrected Sandwich Estimators

Generalized estimating equations (GEE) are often used to analyze Cluster Randomized Clinical Trials (CRTs). When there are few clusters and when cluster sizes are unequal, the GEE analysis can be impacted. The objective of this research is to evaluate how GEE estimates are affected by varying cluster sizes in CRTs with binary outcomes. The degree of variation in cluster size is quantified by the coefficient of variation (CV). With a series of simulation studies, we demonstrate the impact of cluster size imbalance on the empirical distribution of the GEE Wald statistic in different settings. We then propose a CV adjusted GEE Wald t-test to account for varying cluster sizes. This test is compared to the tests recommended by other researchers to preserve the nominal Type I error rate (Li & Redden, 2014). As cluster size variability increases, small clusters become relatively uninformative in the calculation of the GEEsandwich variance estimator and lead to its underestimation. Our approach outperforms the existing recommendation by reducing the increases in the Type I error rate when cluster size variability is moderate to severe (CV=0.5-1.5). Results are confirmed in a completed CRT.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association