Goodness-of-fit tests under dependent censoring
A large number of statistical methods for right-censored data rely on the key assumption of independence between the target and censoring variables. However, this assumption is often violated in real applications and, as reported in several studies, may lead to inconsistent statistical procedures.
In this talk, we propose new goodness-of-fit tests for right-censored data that account for dependence between the target and censoring variables through an Archimedean copula framework. The asymptotic distribution of the proposed test statistics is derived under the null hypothesis, and the consistency of the proposed methodology is also established. Since the direct application of these asymptotic results may be challenging in practice, a bootstrap resampling scheme is developed to approximate the null distribution of the tests. The finite-sample performance of the proposed procedures is evaluated through Monte Carlo simulations, and the methodology is illustrated with a real dataset.
Palabras clave: Archimedean copulas dependent censoring goodness-of-fit nonparametric statistics survival analysis