Power analysis for cluster randomized trials with binary outcomes modeled by generalized linear mixed-effects models

Power analysis for cluster randomized control trials is difficult to perform when a binary response is modeled using the generalized linear mixed-effects model (GLMM). Although methods for clustered binary responses exist such as the generalized estimating equations, they do not apply to the context of GLMM. Also, because popular statistical packages such as R and SAS do not provide correct estimates of parameters for the GLMM for binary responses, Monte Carlo simulation, a popular ad-hoc method for estimating power when the power function is too complex to evaluate analytically or numerically, fails to provide correct power estimates within the current context as well. In this paper, a new approach is developed to estimate power for cluster randomized control trials when a binary response is modeled by the GLMM. The approach is easy to implement and seems to work quite well, as assessed by simulation studies. The approach is illustrated with a real intervention study to reduce suicide reattempt rates among US Veterans.

[1]  Eric R. Ziegel,et al.  Generalized Linear Models , 2002, Technometrics.

[2]  X. Tu,et al.  Applied Categorical and Count Data Analysis , 2012 .

[3]  N Dubin,et al.  Estimation and sample size considerations for clustered binary responses. , 1994, Statistics in medicine.

[4]  Amita K. Manatunga,et al.  Sample Size Estimation in Cluster Randomized Studies with Varying Cluster Size , 2001 .

[5]  Hui Zhang,et al.  Modeling longitudinal binomial responses: implications from two dueling paradigms , 2011 .

[6]  D. Pregibon.  Goodness of Link Tests for Generalized Linear Models , 1980 .

[7]  Jeanne Kowalski,et al.  Modern Applied U-Statistics , 2007 .

[8]  X M Tu,et al.  Power analyses for longitudinal trials and other clustered designs , 2004, Statistics in medicine.

[9]  K. Muller,et al.  Power Calculations for General Linear Multivariate Models Including Repeated Measures Applications. , 1992, Journal of the American Statistical Association.

[10]  S. Zeger,et al.  Longitudinal data analysis using generalized linear models , 1986 .

[11]  P. Diggle.  Analysis of Longitudinal Data , 1995 .

[12]  Shinichi Nakagawa,et al.  Repeatability for Gaussian and non‐Gaussian data: a practical guide for biologists , 2010, Biological reviews of the Cambridge Philosophical Society.

[13]  J Rochon,et al.  Application of GEE procedures for sample size calculations in repeated measures experiments. , 1997, Statistics in medicine.

[14]  X M Tu,et al.  Power analyses for longitudinal study designs with missing data , 2007, Statistics in medicine.

[15]  Jessaca Spybrook,et al.  Optimal Design for Longitudinal and Multilevel Research: Documentation for the "Optimal Design" Software , 2006 .

[16]  J. Kalbfleisch,et al.  A Comparison of Cluster-Specific and Population-Averaged Approaches for Analyzing Correlated Binary Data , 1991 .

[17]  Liang Zhu,et al.  On fitting generalized linear mixed‐effects models for binary responses using different statistical packages , 2011, Statistics in medicine.

[18]  Harvey Goldstein,et al.  Partitioning variation in multilevel models , 2002 .

[19]  S J Pocock,et al.  Repeated measures in clinical trials: analysis using mean summary statistics and its implications for design. , 1992, Statistics in medicine.