In this paper, we describe Bayesian modeling of dependent multivariate surv
ival data using positive stable frailty distributions. A flexible baseline
hazard formulation using a piecewise exponential model with a correlated pr
ior process is used. The estimation of the stable law parameter together wi
th the parameters of the (conditional) proportional hazards model is facili
tated by a modified Gibbs sampling procedure. The methodology is illustrate
d on kidney infection data (McGilchrist and Aisbett, 1991).