An exploration of fixed and random effects selection for longitudinal binary outcomes in the presence of nonignorable dropout

• Published in: Biometrical journal Vol. 55; no. 1; pp. 17 - 37
• Main Authors: Li, Ning, Daniels, Michael J., Li, Gang, Elashoff, Robert M.
• Format: Journal Article
• Language: English
• Published: Weinheim Blackwell Publishing Ltd 01.01.2013; Wiley-VCH; Wiley - VCH Verlag GmbH & Co. KGaA
• Subjects: Applications; Bayes Theorem; Bayesian analysis; Bayesian variable selection; Bias; Biology, psychology, social sciences; Biometry - methods; Brain Ischemia - complications; Dropout; Exact sciences and technology; Female; General topics; Humans; Longitudinal Studies; Male; Mathematics; Medical sciences; Missing data; Model selection; Parametric inference; Plasminogen Activators - therapeutic use; Probability and statistics; Randomized Controlled Trials as Topic; Sciences and techniques of general use; Statistics; Stroke; Stroke - complications; Stroke - drug therapy; Biometrics; Bayes estimation; Mixed distribution; Rank statistic; Prior distribution; Model selection; Bias; Life science; Binary data; Bayesian variable selection; Fixed effect; Parametric method; Statistical method; Missing data; Selection problem; Medical science; Simulation; Binary response; Random effect; Clinical trial; Selection method; Biased estimation; Variable selection; Dropout
• ISSN: 0323-3847; 1521-4036; 1521-4036
• DOI: 10.1002/bimj.201100107

We explore a Bayesian approach to selection of variables that represent fixed and random effects in modeling of longitudinal binary outcomes with missing data caused by dropouts. We show via analytic results for a simple example that nonignorable missing data lead to biased parameter estimates. This bias results in selection of wrong effects asymptotically, which we can confirm via simulations for more complex settings. By jointly modeling the longitudinal binary data with the dropout process that possibly leads to nonignorable missing data, we are able to correct the bias in estimation and selection. Mixture priors with a point mass at zero are used to facilitate variable selection. We illustrate the proposed approach using a clinical trial for acute ischemic stroke.