Nonparametric Methods for the Comparison of ROC Curves with Application to Biomedicine

Abstract

The main goal of this project is to use nonparametric methods to obtain new procedures concerning the ROC curve with significant applications in the biomedical field. Firstly, tests for the comparison of ROC curves without covariates will be studied. Next, new tests will be designed and studied for the comparison of ROC curves in the presence of a unidimensional covariate. Eventually, those tests will be adapted to the case in which the covariate is multidimensional. It will also be of great interest to study whether it is necessary to include those covariates in an ROC curve analysis. The new procedures will be developed in R, and later applied to real data provided by the Hospital Universitario de Santiago de Compostela.