Adaptive kernel estimation of the hazard rate function

Abstract EN

In this paper, we generalized the constant bandwidth kernel estimator of the hazard rate function from Watson and Lead better (1964), which depends on a single bandwidth to the adaptive kernel estimator, which depends on different bandwidths.

We derive the asymptotic normality of the adaptive kernel estimator.

Also we illustrate the use of the adaptive kernel hazard rate estimator in both simulation and real life data and compared it to the constant bandwidth kernel estimator.

In our study, we show that the adaptive estimator has no boundary effects as the constant bandwidth kernel estimator, and has smaller bias.