MAXIMUM-LIKELIHOOD-ESTIMATION OF RECEIVER OPERATING CHARACTERISTIC (ROC) CURVES FROM CONTINUOUSLY-DISTRIBUTED DATA

We show that truth-state runs in rank-ordered data constitute a natura l categorization of continuously-distributed test results for maximum likelihood (ML) estimation of ROC curves. On this basis, we develop tw o new algorithms for fitting binormal ROC curves to continuously-distr ibuted data: a true ML algorithm (LABROC4) and a quasi-ML algorithm (L ABROC5) that requires substantially less computation with large data s ets. Simulation studies indicate that both algorithms produce reliable estimates of the binormal ROC curve parameters a and b, the ROC-area index A(z), and the standard errors of those estimates. (C) 1998 John Wiley & Sons, Ltd.