Gaur-gaurkoak
Ekitaldi eta jardunaldiak
Data: 2024(e)ko martxoaren 14a 10:00
Seminario de investigación: Is AUC a misleading measure of the performance of a biomarker?
Por Mª del Carmen Pardo. Universidad Complutense de Madrid
INAMAT2 organiza este seminario que se celebrará el día 14 de marzo a las 10:30 en la sala de conferencias del edificio Jerónimo de Ayanz.
ABSTRACT:
The area under the receiver operating characteristic (ROC) curve, known as the AUC, is currently considered to be the standard method to assess the discriminatory ability of a marker. In this talk, we review some of the features of this measure and bring into question its reliability as a measure of accuracy for classifier evaluation. As alternative, we explore three different approaches: new summary ROC indexes such as the length of the ROC curve, measures of distance between distributions and overlap measures. To obtain new summary ROC measures based on measures of distance between distributions, we shall exploit the relationship between two-sample test statistics and ROC summary indexes in a similar way as AUC and the Mann-Whitney U-statistic. Thirdly, overlap measures, which capture the similarity between two distributions are considered for determining the diagnostic performance of biomarkers. Furthermore, we identify situations in which these new proposals outperform AUC through several simulation studies. Finally, the procedures will be illustrated via the analysis of real data sets.