Abstract: We prove the identifiability property of Archimedean copula models for dependent competing risks data when at least one of the failure times is exponentially distributed. With this property, it becomes possible to quantify the dependence between competing events based on exponentially distributed dependent censored data. We demonstrate our estimation procedure using simulation studies and in an application to survival data.

Key words and phrases: Archimedean copula models, copula graphic estimator, identifiability of competing risks data.