Comparison of receiver operating curves derived from the same population: a bootstrapping approach

摘要: Abstract The receiver operating curve (ROC) gives a representation of sensitivity and specificity prediction model when varying the cutpoint decision rule on whole spectrum. Evaluation two models established (or tested) in same population patients warrants valid statistical comparison their ROC curves. Hanley et al. recently provided method for overall curves (J. A. B. J. McNeil, Radiology 148 , 839–843, 1983). Often cross, or differ only part courses. Bootstrapping is proposed as graphical check significance differences confined to curve. An example comparing coronary artery disease progression given illustrate this new approach.