Energy distance-based goodness-of-fit test for randomly right censored data

M. Vidal-García, R. M. Crujeiras, W. González-Manteiga

When constructing goodness-of-fit tests, a well-established nonparametric approach is to use some kind of distance between distributions. The choice of a specific weighted L2 distance leads to a particular measure closely connected with energy distance. Due to its interesting properties, a wide variety of energy distance-based tests have been proposed such as test of independence, test for comparing populations or goodness-of-fit test.

In Survival Analysis there are usually restrictions when observing the variable of interest (survival times) in a sample, so collected data will combine complete and partial observations. A particularly popular scenario is that of random right censoring. Previously mentioned energy distance-based test of independence has already been adapted to this context.

In this talk, we will present a new energy distance-based goodness-of-fit test for right censored data. The performance of the test will be illustrated via simulations and real data application.

Palabras clave: energy distance, goodness-of-fit, randomly right censored data, survival analysis

GT09.NOPAR4 Sesión invitada. Contrastes no paramétricos
9 de noviembre de 2023  15:30
CC1: Auditorio

A. Fanjul Hevia, W. González Manteiga, J. C. Pardo Fernández

A. Fernández de Marcos Giménez de los Galanes, E. García Portugués

J. de Uña Álvarez, M. D. Jiménez Gamero