The effects of reordering the response categories on the area under the ROC

The area under the rating ROC can be a useful index of stimulus discriminability. It has the advantage that few assumptions must be made about the underlying distributions of signal and noise. One of the assumptions that must be made is the order of the subject’s response scale. A procedure is outlined for determining the order of the response scale actually used by the subject, and some of the implications of reordering the scale used in analysis are discussed. Results of computer simulation of the effects of varying some important experimental parameters are presented.