Non-parametric estimation of ROC curve

Abstract

Receiver operating characteristic (ROC) curve is widely applied in measuring discriminatory ability of diagnostic or prognostic tests. This makes ROC analysis one of the most actively research areas in medical statistics. Many parametric and nonparametric estimation methods have been proposed for estimating the ROC curve and its functionals. In this paper, we introduce a non-parametric method based on the Bayesian bootstrap technique to estimate ROC curves for continuous diagnostic variables based on independent observations. The area under the ROC curve (AUC) is used to measure the accuracy of different diagnostic methods. The accuracy of the estimate of the ROC curve in the simulation studies is examined by the integrated absolute error (IAE). In comparison with other existing curve estimation methods, the Bayesian bootstrap method compares favorably in terms of accuracy, robustness and simplicity.