BAYESIAN ANALYSIS OF UNEMPLOYMENT DURATION DATA IN THE PRESENCE OF RIGHT AND INTERVAL CENSORING
In this paper, Bayesian inference for unemployment duration data in the presence of right and interval censoring, where the proportionality assumption does not hold, is discussed. In order to model these kinds of duration data with some explanatory variables, Bayesian loglogistic, log-normal and Weibull accelerated failure time (AFT) models are used. In these models, sampling from the joint posterior distribution of the unknown quantities of interest are obtained through the use of Markov chain Monte Carlo (MCMC) methods using the available WinBUGS software. These models are also applied for unemployment duration data of Iran in 2009. The models are compared using deviance information criterion (DIC). Two new sensitivity analyses are also performed to detect: (1) the modification of the parameter estimates with respect to the alteration of generalized variance of the multivariate prior distribution of regression coefficients, and (2) the change of the posterior estimates with respect to the deletion of individuals with high censoring values using Kullback-Leibler divergence measure.