Statistics and Its Interface
Volume 12 (2019)
Bayesian approach for clustered interval-censored data with time-varying covariate effects
Pages: 457 – 465
Interval-censored data arise when the failure time cannot be observed exactly but can only be determined to lie within an interval. Interval-censored data are very common in clinical trials and epidemiological studies. In this study, we consider a Bayesian approach for clustered interval-censored data under a dynamic Cox regression model. Some methods that incorporate right censoring have been developed for clustered data with temporal covariate effects. However, interval-censored data analysis under the same circumstance is much less developed. In this paper, we estimate piecewise constant coefficients based on a dynamic Cox regression model under the Bayesian framework. The dimensions of coefficients are automatically determined by the reversible jump Markov chain Monte Carlo algorithm. Meanwhile, we use a shared frailty factor for unobserved heterogeneity or for statistical dependence between observations. Simulation studies are conducted to evaluate the performance of the proposed method. The methodology is exemplified with a pediatric study on children’s dental health data.
Cox model, frailty, interval censoring, reversible jump Markov chain Monte Carlo, time-varying coefficient, children’s dental health data
Dr. Yue Zhang was supported by the Shanghai Philosophy and Social Sciences Planning Project (2018EJB006) and the Fundamental Research Funds for the Central Universities (17X100040066).
Received 22 June 2017
Published 4 June 2019