|
Abstract:
|
A class of joint dynamic models are proposed to simultaneously model recurrent competing risks and a terminal event (RCRTE) by Liu and Peña (2016). The joint dynamic models include both no frailty and frailty cases. The latter case introduces a frailty variable Z that further induces associations between recurrent competing risks and the association between a recurrent competing risk and the terminal event. For the recurrent competing risks, the intensity processes of the counting processes take into account past event occurrences and impact of possible interventions through effective age processes. For the terminal event, the model also considers impact of past recurrent competing events. In this paper, we propose a model diagnostics procedure that utilizes martingale residuals to assess model fit. The paper also addresses issues of model misspecification and discusses consequences of model misspecification on the estimators proposed in RCRTE.
|
