Semiparametric estimation of single-index hazard functions without proportional hazards

This research develops a semiparametric kernel-based estimator of hazard functions which does not assume proportional hazards. The maintained assumption is that the hazard functions depend on regressors only through a linear index. The estimator permits both discrete and continuous regressors, both discrete and continuous failure times, and can be applied to right-censored data and to multiple-risks data, in which case the hazard functions are risk-specific. The estimator is root- n consistent and asymptotically normally distributed. The estimator performs well in Monte Carlo experiments.