Power Under Local Alternatives for Generalized Estimating Equations

• Activity Number: 114
• Type: Topic Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Zhigang Li*+
• Companies: Columbia University
• Keywords: Asymptotic distribution ; GEE ; Noncentral chi-square ; Power ; Sample size

Abstract

We consider the problem of calculating power based on the GEE method in biomedical studies involving longitudinal or clustered data. Existing approaches approximate the power based on fixed alternatives. A more rigorous approach is to study the power based on a sequence of alternatives that converge at root-n-rate to the null hypothesis. Our contribution is to develop a method of power analysis for GEE based on Wald and quasi-score statistics under such alternatives. Our results show that existing approach under marginal models produces reliable power calculations for linear regression models. However, in the important special case of the Wald test statistic in logistic regression with exchangeable correlation structure, existing method under marginal models inflates the the projected sample size (to obtain nominal 90% power) by about 10% in various cases.