Choose Language Hide Translation Bar
Highlighted
Harj
Level II

Generating confidence intervals for sensitivity, specificity and AUC in ROC curves

Hi,

Can anyone help me with how to generate 95% confidence intervals in ROC curve analyses?

 

I can generate the curves and view the tables but cannot find an easy way of getting the confidence intervals for sensitivity, specificity and AUC.

 

Many thanks. 

 

 

1 ACCEPTED SOLUTION

Accepted Solutions
Highlighted
KarenC
Super User

Re: Generating confidence intervals for sensitivity, specificity and AUC in ROC curves

There are a couple of ways to get to what you need:

 

1. If you have JMP Pro then you can use one-click bootstrapping to get a bootstrap CI for AUC.

2. The add-in https://community.jmp.com/t5/JMP-Add-Ins/Exploring-Model-Classification-Thresholds/ta-p/66964 will provide AUC CI and Se and Sp CI and more.

3. The add-in https://community.jmp.com/t5/JMP-Add-Ins/Performance-Summary-for-Diagnostic-Tests/ta-p/22524 is a simpler add-in that just does the performance measures portion of the add-in in #2.

4. You can always get the Se (or Sp) CI using the distribution platform using summary table with a column for 0/1 and a column for counts of each and then use distribution with column 1 (a nominal column) as the Y and the column of counts as the Freq then select CI from the red triangle menu (you can use Redo>Automatic Recalc to look at different results).

 

 

View solution in original post

1 REPLY 1
Highlighted
KarenC
Super User

Re: Generating confidence intervals for sensitivity, specificity and AUC in ROC curves

There are a couple of ways to get to what you need:

 

1. If you have JMP Pro then you can use one-click bootstrapping to get a bootstrap CI for AUC.

2. The add-in https://community.jmp.com/t5/JMP-Add-Ins/Exploring-Model-Classification-Thresholds/ta-p/66964 will provide AUC CI and Se and Sp CI and more.

3. The add-in https://community.jmp.com/t5/JMP-Add-Ins/Performance-Summary-for-Diagnostic-Tests/ta-p/22524 is a simpler add-in that just does the performance measures portion of the add-in in #2.

4. You can always get the Se (or Sp) CI using the distribution platform using summary table with a column for 0/1 and a column for counts of each and then use distribution with column 1 (a nominal column) as the Y and the column of counts as the Freq then select CI from the red triangle menu (you can use Redo>Automatic Recalc to look at different results).

 

 

View solution in original post

Article Labels

    There are no labels assigned to this post.