Semiparametric competing risks analysis

摘要

In this paper we analyse a semi-parametric estimation technique for competing risks models based on series expansion of the joint density of the unobserved heterogeneity components. This technique allows for unrestricted correlation among the risks. The finite sample behavior of the estimation technique is analysed in a Monte Carlo experiment using an empirically relevant data-generating process. The estimator performs well when compared with the Heckman-Singer estimator.