Dear Gerald
Power calculations for ROC analysis can be pretty meaningless -
ROC curves are descriptive tools, and there are several ways an
"area difference of 5% can arise". ROC analysis is also very
dependent on normality assumptions, and power calculations would
differ depending on the use of parametric or nonparametric ROC
analysis. How do you decide whether a 5% area difference is "clinically
significant"?
Nevertheless, see STATISTICS IN MEDICINE, VOL. 16, 1529–1542
(1997) SAMPLE SIZE DETERMINATION FOR DIAGNOSTIC ACCURACY STUDIES
INVOLVING BINORMAL ROC CURVE INDICES by NANCY OBUCHOWSKI
Aubrey Blumsohn
GM> I am planning a prospective evaluation of a new test and will compare its
GM> performance with that of the test it will replace.
GM> Any advice would be appreciated in estimating the number of patients needed
GM> to be able to detect a clinically significant difference (say 5 or 10%)
GM> between the areas under two ROCs.
GM> Gerald Maguire
GM> Deparment of Clinical Biochemistry and Immunology
GM> Box 232, Addenbrooke's Hospital
GM> Cambridge, CB2 2QR
------ACB discussion List Information--------
This is an open discussion list for the academic and clinical
community working in clinical biochemistry.
Please note, archived messages are public and can be viewed
via the internet. Views expressed are those of the individual and
they are responsible for all message content.
ACB Web Site
http://www.acb.org.uk
List Archives
http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
List Instructions (How to leave etc.)
http://www.jiscmail.ac.uk/
|