Confidence interval for Diagnostic Accuracy (sensitivity and specificity)

Emmanuel_Romeu

Hello,

is there a way to get the Confidence interval for Diagnostic Accuracy (sensitivity and specificity) ?

When using the contingency table I can get the Specificity (%column of TN i.e. true negative), the Spensitivity (%column of TP i.e. true positive) but a customer need the 95% CI using the Wislon score method 

https://www.itl.nist.gov/div898/handbook/prc/section2/prc241.htm

as recomended by FDA

karencopeland

Emmanuel,

 

You can get this using distribution...make a table with column 1 = 0/1, nominal and column 2 = counts then distribution with freq, then pull up the CI (these are score CI)....after getting tired of doing this many times I wrote an add-in Performance Summary Add-in.

There was also a summer intern that wrote an add -in that does even more  https://community.jmp.com/t5/JMP-Add-Ins/Exploring-Model-Classification-Thresholds/ta-p/66964.  This one has some performance issues - good new is that much of the threshold exploration is going to be in 16 in model screening and model compare.

 

Karen