Sensitvity, specificity, positive predictive value, negative predictive value

GS2 — Mon, 15 Feb 2021 16:54:16 GMT

Using SAS 9.4

I have a 3 level response variable and continuous predictor variable (see data example below). What would be the best method to get the sensitivity, specificity, positive predictive value and negative predictive value and to create a ROC/AUC plot?

data have;
infile datalines delimiter=',';
input flex $ motor;
datalines;
Baseline,3
Baseline,4
Baseline,5
Baseline,6
Negative, 7
Negative, 8
Negative, 9
Negative, 10
Negative, 5
Negative, 4
Negative, 8
Negative, 9
Positive,10
Positive,7
Positive,3
Positive,5
Positive,6
Positive,8
Positive,10
Positive,4
Positive,3
;

I tried the code below; however, I get the the following note in the log

"NOTE: Since there are more than 2 response levels, the OUTROC= option has no effect."

ods graphics on;
proc logistic data=want outest=parms(keep=intercept motor);
model flex(event='Positive')= motor / outroc=roc1;
output out=out p=phat;
run;
ods graphics off;

Re: Sensitvity, specificity, positive predictive value, negative predictive value

StatDave — Mon, 15 Feb 2021 21:07:37 GMT

The statistics you mention, including the ROC curve, is only defined for a binary response. However, a multinomial version of the AUC statistic is available from the MultAUC macro.

topic Sensitvity, specificity, positive predictive value, negative predictive value in Statistical Procedures

Sensitvity, specificity, positive predictive value, negative predictive value

Re: Sensitvity, specificity, positive predictive value, negative predictive value