Comparison of presmoothing methods in nonparametric latency estimation

S. Saavedra Martínez, A. López Cheda, M. A. Jácome Pumar

Usual nonparametric estimators in survival analysis present a drawback: they are computed giving weight only to the uncensored observations, leading to efficiency loss specially in the presence of high censoring. Presmoothing techniques (Cao and Jácome, 2004) overcome this inconvenience by assigning non-zero weights to the censored observations through modified weighting assignments.
In cure models, which study the time to event under the assumption that a fraction of the population will never experience the event, presmoothing was applied for the estimation of the cure rate (Saavedra et al., 2025). Similarly, we propose to improve the latency estimator in López-Cheda et al. (2017) using presmoothing ideas. The present work compares two alternative presmoothing methods, showing their performance with respect to the classical nonpresmoothed estimator.

Palabras clave: Censored data, Cure models, Nonparametric estimation, Presmoothing, Survival analysis

GT Bioestadística SEIO - BIOSTATNET: Análisis de Supervivencia
2 de septiembre de 2026  15:30
Aula B

D. Antonio, F. Blanco, M. Saez Zafra, M. A. Barceló

J. Gutiérrez Botella, C. Armero i Cervera, T. Kneib, J. García-Seara

G. Gómez Melis, I. Arostegui Madariaga, K. Langohr

A. Alcaraz Peñalba, A. M. Franco Pereira, M. C. Pardo Llorente