Validation croisée pour l'estimateur lissé de la fonction de hasard : cas des données censurées
Résumé
In the setting of censored data, we consider an estimator of hazard rate built with a
kemel estimator of the "survival" function of the censored variable. In a first attempt we give
the asymptotic expression of the Integrated Square Error (ISE), in a second attempt we propose
a method based on cross-validation ideas to select the bandwidth; the selected bandwidth is
shown to be asymptotically optimal. A small scale simulation completes this paper.