Dipartimento di Scienze Statistiche

  PhD program in Statistics DSS Statistics Seminar    January 23, 2026, 12:00   In person.Room 34 (CU002)  Webinar^.https://uniroma1.zoom.us/j/83625004899?pwd=bXCtz0 mp759PUh2lkqT0BUoVa0Uegg.1 Passcode: 123456    Developments in Receiver Operating Characteristic Curve Analys       Musie Ghebremichael Harvard Medical School, Ragon Institute of MGH, MIT, and Harvard; and MGH Biostatistics Center    Receiver operating characteristic (ROC) curve analysis is a cornerstone of diagnostic and biomarker evaluation in medical research. As its use has expanded, a wide range of ROC estimation and inference methods — parametric, nonparametric, semiparametric, Bayesian, and placement value– based — have been developed, each relying on different distributional assumptions and yielding curves with distinct properties and area-under-thecurve estimates. This talk reviews recent developments in ROC methodology and compares the performance of these approaches through extensive simulation studies across varying distributional forms, sample sizes, and levels of overlap between diseased and non-diseased populations. Applications to HIV/AIDS data are presented to illustrate practical differences among methods in real-world settings. Results demonstrate that ROC estimates and inferences can vary substantially depending on the chosen methodology, underscoring the importance of careful method selection and validation. The talk concludes with practical guidance for statisticians and applied researchers on selecting ROC inference methods that are well-suited to their data and study objectives.   In allegato la locandina