Reliable and computationally efficient maximum-likelihood estimation of "proper" binormal ROC curves.

摘要: Rationale and Objectives Estimation of ROC curves their associated indices from experimental data can be problematic, especially in multireader, multicase (MRMC) observer studies. Wilcoxon estimates area under the curve (AUC) strongly biased with categorical data, whereas conventional binormal curve-fitting model may produce unrealistic fits. The “proper” (PBM) was introduced by Metz Pan to provide acceptable fits for both sturdy problematic datasets, but other investigators found that its first software implementation numerically unstable some situations. Therefore, we created an entirely new algorithm implement PBM. Materials Methods This paper describes detail PBM algorithm, which designed perform successfully all situations encountered previously. Extensive testing conducted also on a broad variety simulated real datasets. Windows, Linux, Apple Macintosh OS X versions are available online at http://xray.bsd.uchicago.edu/krl/ . Results Plots fitted as well summaries AUC standard errors reported. never failed converge produced good several million datasets it tested. For most very error. compared continuously distributed expected superior data. Conclusion is reliable wide tasks.