Generalisations of the Receiver Operating Characteristic (ROC) Curve

Thursday, 23 July 2015
9.30am - 10.30am
Room 515, Level 5, Melbourne School of Population & Global Health, Melbourne University
207 Bouverie St
Carlton 3052

The ROC curve is a popular graphical method used to study the diagnostic capacity of biomarkers. In its simplest form it plots true-positive rates against false-positive rates. Both practical and theoretical aspects of the properties of ROC curves have been extensively studied. Conventionally, it is assumed that the considered marker has a monotone relationship with the studied characteristic, that is, the upper (lower) values of the biomarker are associated with a higher (lower) probability of a positive result. In many real situations, however, both the lower AND the upper values of the marker are associated with higher probability of a positive result. We propose an ROC curve generalization which is useful in this context. All pairs of possible cut-off points, one for the lower and another one for the upper biomarker value, are considered and the best one is selected. Moreover, the empirical estimator for the curve is considered and its uniform consistence and asymptotic distribution derived. The theoretical framework we provide allow us to derive ROC curve generalizations for multivariate biomarkers.

Dr Pablo Martinez-Camblor

Dr Pablo Martinez-Camblor

Head of Clinical Statistics
Hospital Universitario Central de Asturias

Dr Pablo Martinez-Camblor is Head of Clinical Statistics at the Hospital Universitario Central de Asturias (HUCA) Oviedo, Asturies, Spain, and holds academic appointments as Associate Researcher at the Universidad Autonoma de Chile and Assistant Professor Department of Statistics, Oviedo University, Asturies, Spain. His main research interest is the comparison of diagnostic methods, reflected in a series of methodological articles proposing multiple extensions for ROC curve analysis. In addition to his work on diagnostic tests, he has contributed to research on time-dependent processes and competing risk events.

PDF icon Powerpoint presentation 887.69 KB