On independence testing in the presence of cure data

Abstract

The focus is on the problem of testing independence between two randomly right-censored variables in the presence of cure data, modeled through an additive mixture framework with a cured fraction. Several classes of test statistics are proposed, inspired by Kendall’s tau and its extensions. Their asymptotic properties are derived, and their finite-sample performance is thoroughly investigated. Robustness to different censoring models is also explored. In addition, a potential generalization that can be used in more dimensional settings is presented. A comprehensive simulation study demonstrates the strong competitiveness of the proposed procedures.