In this paper, we propose an empirical likelihood ratio method for the inference about average derivatives in semiparametric hazard regression models for competing risks data. Empirical loglikelihood ratio for the vector of the average derivatives of a hazard regression function is defined and shown to be asymptotically chi-squared with degrees of freedom equal to the dimension of covariate vector. Monte Carlo simulation studies are presented to compare the empirical likelihood ratio method with the normal-approximation-based method.
- Average derivative
- Competing risks data
- Empirical likelihood
- Nonparametric hazard regression
- Single-index model