Robustness to baseline-hazard misspecification

Assess the robustness of the simultaneous Bayesian joint modeling framework and its observed gains over the two-stage strategy when the assumed Weibull baseline hazard is misspecified.

Background

The simulation study compares a frequentist two-stage strategy using a semiparametric Cox model with the simultaneous Bayesian JMBB model, which assumes a Weibull baseline hazard. Because the data are generated under the Weibull specification, the reported advantages of JMBB may reflect not only simultaneous estimation but also differences in baseline-hazard treatment and model specification. The robustness of those gains under alternative or misspecified baseline hazards is therefore left unresolved.