ROCToolbox: a Matlab Toolbox for ROC analysis in neuroinformatics.
Última modificación: 2011-03-26 11:41
Resumen
Receiver operating characteristic (ROC) analysis allows estimating and comparing the accuracy of diagnostic tests for different gold standard type. All possible combinations of sensitivity and specificity achieved by changing the test's cut-off value can be summarized using a single parameter: the area under the ROC curve. The ROC technique can also be used to optimize cut-off values with regard to a given prevalence in the target population and cost ratio of false-positive and false-negative results. However, plots of optimization parameters against the selected cut-off value provide a more direct method for cut-off selection. Here we introduce the ROCToolbox, a Matlab toolbox that implements nonparametric estimators proposed by Obuchowski in 2005 for estimating and comparing diagnostic tests' accuracies when the gold standard is measured on a binary, continuous, ordinal or nominal scale. ROCToolbox also includes some methods for cut-off selection such as linear combinations of sensitivity and specificity, odds ratio and chance-corrected measures of association (e.g. kappa).