Nonparametric hazard function estimation

Nonparametric hazard estimation is investigated by the kernel method. Complete convergence results (with rates) are given and further data driven bandwidths selection rules are introduced and shown to be asymptotically optimal. All the results are stated in cases when the observed random variables are not necessarily independent. As a by-product the authors give also extensions of the Glivenko-Cantelli theorem for random variables satisfying various mixing conditions.