topic Sensitivity and specificity with CI 95% using nominal variable with JMP 13 in Discussions

Sensitivity and specificity with CI 95% using nominal variable with JMP 13

Titanescu — Sat, 25 Apr 2020 09:04:23 GMT

Dear Members of JMP community,

I have JMP 13

I'm attending to assess sensitivity and specificity (with CI 95%) of a nominal variable for another nominal variable.

I'm not able to perform this

May you indicate how to proceed?

Re: Sensitivity and specificity with CI 95% using nominal variable with JMP 13

Mark_Bailey — Sat, 25 Apr 2020 12:57:50 GMT

So you have the confusion matrix, not the original data (observation pairs)? That is, you have a table of true levels versus predicted levels? If so, then JMP has no built-in analysis to help you. It works with the data, not a summary of the data. You would need a script to calculate the sensitivity and specificity. This calculation is simple from the confusion matrix.

Is it a 2x2 table?

Re: Sensitivity and specificity with CI 95% using nominal variable with JMP 13

Titanescu — Sat, 25 Apr 2020 21:06:23 GMT

Thank you very much

I have the original data; for each test perform I have the value of the
nominal variable 1 (gold standard: "control" or "disease") and the value of
the nominal variable 2 (test result : "postive" or "negative" )

This look like:

Gold standard Test result
Test 1 Control Negative
Test 2 Control Negative
Test 3 Disease Positive
Test 4 ect....

Sorry for not having mentionned this at the stage of my first message

Thank you again for your help,

With my best regards

Alan

Re: Sensitivity and specificity with CI 95% using nominal variable with JMP 13

Mark_Bailey — Sun, 26 Apr 2020 12:14:05 GMT

You can get the confusion matrix this way:

Be sure that the Gold Standard and Test Result data columns are using the Nominal modeling type.
Select Analyze > Fit Model.
Select the Test Result column and click Y.
Select the Gold Standard column and click Add.
Click Run.
Click the red triangle at the top and select Confusion Matrix.

You can perform the calculations for both point estimates and interval estimates of the sensitivity and specificity based on this article.