A single risk approach to the semiparametric copula competing risks model

TL;DR

This paper develops a new approach for analyzing competing risks where one risk is modeled parametrically or semi-parametrically, ensuring identifiability and providing estimation methods with good finite sample performance.

Contribution

It introduces a single risk approach for semiparametric copula competing risks models, establishing identifiability and proposing $\

Findings

Identifiability established for popular parametric and semiparametric models.

Proposed estimators are $\

Demonstrated effective through simulations and real data examples.

Abstract

A typical situation in competing risks analysis is that the researcher is only interested in a subset of risks. This paper considers a depending competing risks model with the distribution of one risk being a parametric or semi-parametric model, while the model for the other risks being unknown. Identifiability is shown for popular classes of parametric models and the semiparametric proportional hazards model. The identifiability of the parametric models does not require a covariate, while the semiparametric model requires at least one. Estimation approaches are suggested which are shown to be $\sqrt{n}$-consistent. Applicability and attractive finite sample performance are demonstrated with the help of simulations and data examples.