nonbinROC: Software for Evaluating Diagnostic Accuracies with Non-Binary Gold Standards

Abstract

ROC analysis is a standard method for estimating and comparing diagnostic tests' accuracies when the gold standard is binary. However, there are many situations when the gold standard is not binary. In these situations, traditional ROC methods applied have lead to biased and uninformative outcomes. This article introduces nonbinROC, software for R that implements nonparametric estimators proposed by Obuchowski (2005) for estimating and comparing diagnostic tests' accuracies when the gold standard is measured on a continuous, ordinal or nominal scale. The results produced from these estimators are interpreted in the same manner as in ROC analysis but are not associated with any ROC curve.