A0432
Title: Assessing significance of covariates in mixture cure models using distance correlation
Authors: Blanca Monroy-Castillo - Universidade da Coruna (Spain) [presenting]
Ingrid Van Keilegom - KU Leuven (Belgium)
M Amalia Jacome - Universidade da Coruna (Spain)
Ricardo Cao - University of Coruna (Spain)
Abstract: One of the challenges in cure models is to test whether a covariate influences the cure rate. Distance correlation is a novel class of multivariate dependence coefficients that offers advantages over classical correlation coefficients: it is applicable to random vectors of arbitrary dimensions, not necessarily equal, and it is zero if and only if the vectors are independent. Distance correlation has been applied in a standard survival model without a cure based on the distance covariance between covariates and survival times. However, to the best of our knowledge, distance correlation has not yet been applied in the presence of a cure fraction. One challenge in dealing with cure survival data is that the cure indicator is only partially observed due to censoring. A method to study the effect of a covariate on the probability of cure using distance correlation is proposed, which overcomes the challenge of handling the missingness of the cure indicator.
