CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Nonparametric methods for the comparison of ROC curves with covariate information Authors:  Aris Fanjul Hevia - Universidad de Oviedo (Spain) [presenting]
Juan-Carlos Pardo-Fernandez - Universidade de Vigo (Spain)
Wenceslao Gonzalez-Manteiga - University of Santiago de Compostela (Spain)
Abstract: The Receiver Operating Characteristic (ROC) curve is a statistical tool that combines the notions of sensitivity and specificity to evaluate the discriminatory capability of a classification problem. Whenever more than one classification procedure is available, the ROC curves may be used for comparing their behaviour. Furthermore, these methods may be affected by the information provided by some covariates, and thus they should be taken into account when comparing the corresponding ROC curves. There are several ways to incorporate the covariates into a ROC curve analysis, the main ones being the conditional ROC curves and the covariate-adjusted ROC curves. We use nonparametric techniques to study these curves with the aim of comparing methods of classification and, at the same time deciding whether the covariate information should be taken into account or not.