Query regarding appropriate ROC curve function

[Omar Zahaf]

Dear all

I am calculating the ability of a diagnostic angle measurement that can differentiate between two different diseases. I am trying to draw the ROC and work out the AUC but not sure what ROC function or code is best.

My reference variable is "disease" and my classifier variable is "angle" (this is a continuous measurement, this is normally distributed parametric data).

My code is this:

rocfit Disease Angle, continuous (3)

rocplot

or is this the correct code:

rocfit Disease Angle, continuous(3) generate (v1)

rocplot

Or is this the correct code:

rocreg Disease Angle, probit

rocregplot, classvars(Angle)

I am also a bit confused about the difference between using just probit and ml(maximum likelihood).




Many many thanks in advance

[Omar Zahaf]

anybody able to help?

[Clyde Schechter]

For the task of estimating the area under the ROC curve and drawing the curve I would do -roctab Disease Angle, graph-. It's simple and it does the job. I would avoid -rocfit- and -rocreg-. Those commands impose modeling constraints on the analysis that may or may not be appropriate for your data.