A Bayesian semiparametric regression model for reliability data using effective age

A new regression model for recurrent events from repairable systems is proposed. The effectiveness of each repair in Kijima models I and II is regressed on repair-specific covariates. By modeling effective age in a flexible way, the model allows a spectrum of heterogeneous repairs besides “good as new” and “good as old” repairs. The density for the baseline hazard is modeled nonparametrically with a tailfree process prior which is centered at Weibull and yet allows substantial data-driven deviations from the centering family. Linearity in the predictors is relaxed using a B-spline transformation. The method is illustrated using simulations as well as two real data analyses.