roccomp comparisons chi2

[Steffi Widner]

Dear all,
I am trying to compare different AUROCs using the roccomp command. To evaluate if for one model the AUROC is statistically different to the next, I plan to use the Chi2 test. After running the test
rocgold NBER p1 p2 p3 p4 p5 I would like to store the pvalue for Chi2 test but have not managed so far. Could you help me with that?
Many thanks,
Steffi

[Andrew Musau]

webuse ct2, clear
rocgold status mod1 mod2 mod3
return list
mat l r(p)
local p1= r(p)[1,1]
local p2= r(p)[1,2]
di "`p1'"   
di "`p2'"

Res.:

. rocgold status mod1 mod2 mod3

-------------------------------------------------------------------------------
                       ROC                                           Bonferroni
                      Area     Std. Err.       chi2    df   Pr>chi2     Pr>chi2
-------------------------------------------------------------------------------
mod1 (standard)     0.8828       0.0317
mod2                0.9302       0.0256      2.3146     1    0.1282      0.2563
mod3                0.9240       0.0241      5.2480     1    0.0220      0.0439
-------------------------------------------------------------------------------

. 
. mat l r(p)

r(p)[1,2]
        mod2       mod3
p  .12816517  .02197149


. di "`p1'"   
.1281651684953875

. 
. di "`p2'"
.0219714862120753

[Rich Goldstein]

your question is not completely clear to me; for example if one model is nested in another, the p-value for the added variable is the p-value for the difference in AUROC's; I am out of town but have previously cited a couple of articles that prove this; use your favorite search engine to look for "Goldstein Pepe site: statalist.org"; but maybe you are asking something different?

[Steffi Widner]

Andrew Musau : this works really well, thanks!
Rich Goldstein : Thanks anyways for the recommendation!