Bayesian random effects selection in mixed accelerated failure time model for interval-censored data

• Published in: Statistics in medicine Vol. 33; no. 6; pp. 971 - 984
• Main Authors: Harun, Nusrat, Cai, Bo
• Format: Journal Article
• Language: English
• Published: England Blackwell Publishing Ltd 15.03.2014; Wiley Subscription Services, Inc
• Subjects: accelerated failure time models; Bayes Theorem; Bayesian analysis; Belgium; Biostatistics; Child; Cholesky decomposition; Computer Simulation; correlated data; Correlation analysis; Data Interpretation, Statistical; Dirichlet process priors; Female; Gaussian mixtures; Humans; Longitudinal Studies; Male; Markov Chains; Medical statistics; Models, Statistical; Monte Carlo Method; Normal distribution; Oral Health - statistics & numerical data; Time Factors; Uncertainty; variable selection; Belgium; Cholesky decomposition; accelerated failure time models; variable selection; Dirichlet process priors; correlated data; Gaussian mixtures
• ISSN: 0277-6715; 1097-0258; 1097-0258
• DOI: 10.1002/sim.6002

In many medical problems that collect multiple observations per subject, the time to an event is often of interest. Sometimes, the occurrence of the event can be recorded at regular intervals leading to interval‐censored data. It is further desirable to obtain the most parsimonious model in order to increase predictive power and to obtain ease of interpretation. Variable selection and often random effects selection in case of clustered data become crucial in such applications. We propose a Bayesian method for random effects selection in mixed effects accelerated failure time (AFT) models. The proposed method relies on the Cholesky decomposition on the random effects covariance matrix and the parameter‐expansion method for the selection of random effects. The Dirichlet prior is used to model the uncertainty in the random effects. The error distribution for the accelerated failure time model has been specified using a Gaussian mixture to allow flexible error density and prediction of the survival and hazard functions. We demonstrate the model using extensive simulations and the Signal Tandmobiel Study®. Copyright © 2013 John Wiley & Sons, Ltd.