Can't graph more than 40 ROC curves using rocgold

[Simon Walsh]

I have a diagnostic reference standard (reference), a test with probabilities of a disease (uip) and then about 90 doctors binary diagnosis of has disease (=1) and has not disease (=0)

* Example generated by -dataex-. To install: ssc install dataex
clear
input float uip byte(reference bin_uip_u1 bin_uip_u2 bin_uip_u3 bin_uip_u4)
1.00e-06 0 0 0 0 0
 .999836 1 1 1 1 1
 .319859 1 1 1 0 0
 .002651 0 0 1 0 0
 .992511 1 1 1 1 1
 .000206 0 0 0 0 0
 .009615 0 0 0 0 0
 .001493 0 0 0 0 0
8.00e-06 0 0 0 0 0
 .001381 0 0 0 0 0

In my example above I have only included up to bin_uip_u4, but they go up to bin_uip_u90

I want to display the a ROC curve for uip, and then show the curves for all the binary variables on the same graph like this.

The difficulty is that it seems to be impossible for more than 42 variables (I am using Stata/IC 14.2 for Mac). Also, I can't seem to automatically remove the lines for the binary variables - I really just want the uip curve, and then the single points for the binary variables. Many thanks in advance for your help.

[Clyde Schechter]

It isn't possible to troubleshoot code that is not shown. The data example is helpfully displayed, but given that your problem is a syntax error message, the problem is far more likely to be with your code.

There is no limit of 40 graphs in -rocgold-. Perhaps there should be, as I don't know how a graph with 40+ curves on it could be anything but an unreadable mess. But there isn't. I am able to create a graph with 50 such curves using -rocgold-.

I suggest that you show your full code, and the complete output (incl. error messages) exactly as they are. Perhaps it will become clear then.

I really just want the uip curve, and then the single points for the binary variables.

I don't understand what you mean here. Please explain. All the variables are named uip*, and all of them, at least in your example, are binary.