Comparing ROC curve

[Kelvin Wan]

I am using the -comproc function to compare two areas under the ROC curve estimated from two independent samples. I want to test whether the sample size has sufficient power to detect a difference. Can anyone point me in the right direction? Thanks.

[Joseph Coveney]

I'm not sure what the "comproc function" is, but you're likely to need to do power analysis for this by simulation. You can get background on how from a StataCorp FAQ and from a related Stata Journal article. You can find step-by-step procedures for ROC simulations in the appendix to this article. See also this on the advisability of what you're trying to do.