S. Saavedra Martínez, A. López Cheda, M. A. Jácome Pumar
In classical survival analysis, the time until an event such as death or relapse in a disease is modelled. It is usually assumed that all individuals will experience it. Cure models emerge to analyse the frequent situations where some individuals will not suffer the event. These individuals are considered cured. One of the main inconvenience in nonparametric kernel methods is that only the uncensored observations are given weights, thus losing precision. Presmoothing techniques solve this problem by giving some weights also to the censored observations. Considering the methodology by Cao and Jácome (2004) to improve the cure rate estimator in López-Cheda et al. (2017), Saavedra et al. (2025) proposed a presmoothed nonparametric estimator for the probability of cure in mixture cure models. In this work, we introduce a bootstrap bandwidth selection method. The performance of the method is evaluated in a simulation study. Finally, the proposed methodology is applied to a real dataset.
Keywords: Bootstrap, Censored data, Cure models, Nonparametric estimation, Presmoothing, Survival analysis.
Scheduled
Non Parametric Statistics II
June 12, 2025 7:00 PM
Auditorio 1. Ricard Vinyes