Bayesian Inference of Interval-Censored Survival Data - Interval-Censored-Time-to-Event Data: Methods and Applications - page 183

• out : name of output file to store the MCMC samples.

•

model : model type to fit. The three options, TimeIndep , TimeVarying ,
and Dynamic , correspond to the Cox proportional hazards model, the
time-varying-coecient Cox model (Sinha et al., 1999), and the dynamic

Cox model (Wang et al., 2011a), respectively.

• base.prior : list of options for prior of baseline hazards.

• coef.prior : list of options for prior of regression coecients.

• gibbs : list of options for Gibbs sampler.

• control : list of general control options.

More detailed documentations and working examples are available at the help

page of function bayesCox in the package.

7.6 Simulation Study

A simulation study was carried out to assess the recovery of true coecients

and the performance of the model selection criteria. We considered a situation

with only a binary covariate X, which follows a Bernoulli distribution with

rate parameter 0.5. Data sets of sample size n = 400 were generated from

model (7.1) with two specications for regression coecient (t): (t) = 1
and (t) = 0:5 + sin(t=6). The baseline hazard function is 0 (t) = 0:1 p (t).
The time interval of interest is set as (0; 6).

Interval-censored survival times were generated in two steps. In the first

step, the exact event time T was generated given X, (t), and 0 (t) by solving
F(T) = U, where F is the distribution function of T, and U is an indepen-

dent uniform variable over (0; 1). In the second step, a censoring interval was

generated. For each subject, we generated gap times as visit schedule until