Bias–corrected methods for estimating the receiver operating characteristic surface of continuous diagnostic tests

摘要

Verification bias is a well-known problem that may occur in the\udevaluation of predictive ability of diagnostic tests. When a binary disease\udstatus is considered, various solutions can be found in the literature to cor-\udrect inference based on usual measures of test accuracy, such as the receiver\udoperating characteristic (ROC) curve or the area underneath. Evaluation of\udthe predictive ability of continuous diagnostic tests in the presence of ver-\udification bias for an ordinal three-class disease status is here discussed. In\udparticular, several verification bias-corrected estimators of the ROC surface\udand of the volume underneath are proposed. Consistency and asymptotic\udnormality of the proposed estimators are established and their finite sample\udbehavior is investigated by means of Monte Carlo simulation studies. Two\udillustrations are also given.