You may find this discussion useful: Use and Misuse of the Receiver
Operating Characteristic
Curve in Risk Prediction - NancyR.Cook,ScD - (Circulation.2007;115:928-935.)
Harrell's c is a roc-like measure. The discussion in the above paper
is informative. You also might want to search for the follow-up papers
and comments. Bottom line: Harrell's c is sometimes, but not always
helpful; it all depends on what is more important to your research.
Sometimes it is HR.
Fred
On Tue, Jan 20, 2009 at 10:33 PM, Zoe Hyde <[email protected]> wrote:
> Dear All,
>
> I have two Cox models that I am using to predict stroke. One model includes total testosterone (HR=2.0) and the other free testosterone (HR=1.7), as Z-scores, along with some other covariates. The hazard ratios seem to suggest that changes in total testosterone are more important than free testosterone, but is this really the case? Is there some way that I can compare the two models, for example by using ROC curves?
>
> I have used estat con to obtain Harrell's C for both models, and there doesn't seem to be a lot of difference - 0.69 for one model and 0.67 for the other. However, I'd feel more comfortable if there was some way I could test the significance of the difference. Is there is test I can use to compare these two results?
>
> Any advice would be much appreciated!
>
>
> Thanks,
>
> Zoe.
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
--
Fred Wolfe
National Data Bank for Rheumatic Diseases
Wichita, Kansas
NDB Office +1 316 263 2125 Ext 0
Research Office +1 316 686 9195
[email protected]
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
