Reader Comments

Post a new comment on this article

Definition of Specificity at Table 4

Posted by nlsim on 08 Sep 2011 at 07:18 GMT


Just wondering if Specificity should be TN / (TN + FP) instead of TP / (TP + FP)? If the former is correct, then the Specificity values for deFuse-Threshold, deFuse-Classifier and FusionSeq should be 4/7, 3/4 and 0 respectively, instead of 100%, 94% and 76.5%...

No competing interests declared.

RE: Definition of Specificity at Table 4

sohrab replied to nlsim on 13 Sep 2011 at 20:48 GMT

Hello nlsim -
Thank you for pointing this out. You are correct that the numbers reported are not specificity. Rather they are positive predictive value (PPV). The Table 4 column header for the last column should therefore instead be PPV. It is also true that the actual specificity values should be: 4/7, 3/4 and 0 respectively as you state in your comment.

No competing interests declared.