Parametric and semiparametric methods often fail to capture the right shape of the conditional hazard rate in survival analysis. In this paper we propose a new and intuitive nonparametric estimator for the conditional hazard rate, based on local linear estimation techniques. This estimator can deal with both censored and uncensored data. We show that the local linear hazard rate estimator is consistent and asymptotically normal distributed. Moreover, we derive plug-in bandwidths based on normal and uniform reference distributions. We show that these bandwidths perform reasonably well, even when the underlying distributional assumptions are violated. We illustrate the use of the nonparametric local linear hazard rate estimator and the bandwidth selection method in several simulation experiments and in two applications to real-life data.
|Publisher||Department of Applied Mathematics, University of Twente|