Testing additivity in nonparametric regression under random censorship

TL;DR

This paper introduces a statistical test for additivity in multivariate nonparametric regression models with right censored data, enabling more accurate modeling in survival analysis and related fields.

Contribution

It proposes a new asymptotically normal statistic specifically designed to test for additivity under censoring conditions.

Findings

The test statistic is asymptotically normally distributed under the additive assumption.

The method effectively detects non-additivity in censored regression models.

Applicable to multivariate right censored data in various statistical analyses.

Abstract

In this paper, we are concerned with nonparametric estimation of the multivariate regression function in the presence of right censored data. More precisely, we propose a statistic that is shown to be asymptotically normally distributed under the additive assumption, and that could be used to test for additivity in the censored regression setting.