Inference with structured hazard functions: from modeling to inference

S. Sperlich, D. Liu, M. Hiabu

Structural modeling in else nonparametric data analysis is today as important as it has been since the beginning of the use of nonparametric methods in data analysis. It is is a way of making models interpretable
and computationally more attractive. The separability of the impacts of covariates in a model is one of the most popular structures to render estimates understandable or to break down the curse of dimensionality. We consider different methods for estimating structured hazard functions, introduce tools for doing further inference where missing, and compare these procedures. Thinking of data sets of moderate size, our main focus is on smoothed backfitting and splines, as more sophisticated machine learning algorithms require large(r) data sets. We further study the possibility of fast implementation, numerical issues that relate to identifyability or comparability, and potential improvements (like local linear versions, partial parametrization, etc.) of the estimators.

Keywords: semiparametric hazard functions, structured semiparametric models, smoothed backfitting, multiplier bootstrap

GT Estadística no Paramétrica IV: Inferencia no paramétrica en análisis de supervivencia
September 5, 2026  4:00 PM
Aula 29

P. Soto Rodríguez, J. de Uña Álvarez, J. C. Pardo Fernández

J. C. Escanciano, J. de Uña Álvarez

W. González-Manteiga, M. D. Martínez-Miranda, I. Van Keilegom, M. Conde Amboage