I would appreciate if you could tell me when to use diagnostic Odds ratio(DOR) instead of AUC (Area under the curve)? or when to have DOR along with AUC ?
Hi, they represent different approaches to the question. Depends how much data you have.
see
Hurley J. Meta-analysis of clinical studies of diagnostic tests: developments in how the receiver operating characteristic “works”. Archives of pathology & laboratory medicine. 2011 Dec 1;135(12):1585-90.
context an data dependent where odds ratio OR=stat1/stat2 and preferably logged(OR) to shift the statistic to a domain of -inf and inf. Under reasonable sample size logged(OR) can be treated as behaving approx normal. This favorable to derive the standard error (often via walds approach). For the AUC this becomes more tricky and while one can derive standard error it is more tricky although one could transform the AUC to ratios the more we transform the more wonky it becomes. The AUC more an overlapping criteria. Then again the context is unclear to me.
I wrotte an article on this - "Hurley JC. (2011) Meta-analysis of studies of diagnostic tests: developments in how the ROC ‘works’ Arch. Pathol Lab Med. 135;;1585-1590. http://www.archivesofpathology.org/doi/abs/10.5858/arpa.2011-0016-SO"