The nonparametric ROC area is essentially the Wilcoxon-Mann-Whitney linear rank
statistic. A book on linear rank statistics will give you all you need.
FE Harrell
On Wed, 27 Aug 2003 08:17:47 -0700 (PDT)
Xao Ping <xao_ping@yahoo.com> wrote:
> Dear All:
> Many thanks to Seen Keenan, Brad Biggerstaff, Michael Westphal, Gerrrit
> Eincher, Frank Harrel and Beth Atkinson for the help with my request about
> the ROC curves.
> After careful study of all the information that was brought to my attention,
> I successfully wrote my own simple function for calculation of ROC curves,
> AUC and SE(AUC).
>
> However, I have another ROC question, probably much more complicated. Is that
> possible to theoretically derive the distribution of AUC, or at least its
> first four cumulants?
> If not, then would it be correct to claim that because the AUC is a function
> of order statistics, then the distribution of AUC is independent, or at least
> highly conservative, with respect to parent distributions for cases and
> controls?
> Any references/thoughts are greatly appreciated.
>
> Thanks again for help
> Xao Ping
> R&R Pharmakinetics
> Taiwan
>
>
>
> ---------------------------------
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site design software
---
Frank E Harrell Jr Prof. of Biostatistics & Statistics
Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences
U. Virginia School of Medicine http://hesweb1.med.virginia.edu/biostat
|