Sensitivity and specificity

2015-01-13

azim58 - Sensitivity and specificity


sensitivity is true positive rate: the proportion of actual positives which are correctly identified as such
specificity is true negative rate: proportion of negatives which are correctly identified as such

Sensitivity = TP/(TP+FN)
Specificity = TN/(TN+FP)

Table
Condition: Condition Positive Condition Negative
Test Outcome
Positive TP FP
Negative FN TN


Mneumonics for sensitivity and specificity
http://books.google.com/books?id=HW0hr7c5l8MC&pg=PA219&lpg=PA219&am
p;dq=how+can+i+remember+the+difference+between+sensitivity+and+specificity&
amp;source=bl&ots=pHYR7sL1RX&sig=go-9Hy7IARwHVWCgYREluHTHmG8&hl
=en&sa=X&ei=5iQEUeydE6XUiwLdnIDQAw&ved=0CEEQ6AEwAQv=onepage&am
p;q=how%20can%20i%20remember%20the%20difference%20between%20sensitivity%20a
nd%20specificity&f=false
http://forums.studentdoctor.net/archive/index.php/t-125098.html


===========================================================================
SeNsitive tests are Negative in health (NIH)
SPecific tests are Positive in Disease (PID)

===========================================================================
Sensitivity: SnOut- sensitive tests rule out diesease - meaning that on a
test with high sensitivity if you test negative chances are you don't
have the disease (screening). If you rule disease out this means you have
few false negatives.
Specificity: SpIn- specific tests rule in disease - meaning that on a
test w. high specificity if you test positive chances are you got the
disease (confirmatory). If you rule disease in this means you have few
false positives.

===========================================================================
SeNsitivity includes False Negatives
SPecificity includes False Positives




===========================================================================

===========================================================================
Note that a ROC curve has true positives vs false positives
(1-specificity)


nice sensitivity specificity article with some examples