0
$\begingroup$

Is there any tool that can calculate the AUC value from a ROC curve if I already have true positive, true negative, false positive, false negative values.

$\endgroup$
6
  • 5
    $\begingroup$ You will need those 4 values at a variety of thresholds. $\endgroup$
    – Zach
    Commented Jul 31, 2013 at 17:32
  • $\begingroup$ You might want to consider just using some descriptive stats in your case, pos. predictive, recall, f1 score, or something similar. As @Zach points out there is no ROC curve when you only have a 2x2 table. $\endgroup$
    – Max Gordon
    Commented Jul 31, 2013 at 20:35
  • $\begingroup$ @MaxGordon none of those values are necessary to calculate the AUC (or $c$-statistic). You only need estimates of sensitivty and specificity obtained from thresholding a probabilistic marker across all possible ranges. $\endgroup$
    – AdamO
    Commented Jul 31, 2013 at 21:27
  • $\begingroup$ @AdamO Those 4 values are used to calculate sensitivity and specificity. $\endgroup$
    – Zach
    Commented Jul 31, 2013 at 23:03
  • $\begingroup$ @Adamo I'm aware of that, it was just a suggestion for what can be done with a simple 2x2 table. As there is no range in a 2x2 table the c-statistic is unattainable as Zach pointed out in the first comment. $\endgroup$
    – Max Gordon
    Commented Aug 1, 2013 at 7:13

0