multiple ROC curves, logistic regression

[seol]

Hi anagrammarian

Yesterday, In psychoPDA > Test ROC module, I noticed that the deLong Test does not resolve the missing value and gives an error, so I am adding a deLong Test that resolves the missing value issue to snowCluster > ROC Analysis.

I hope this will help you.

Best Regards
Seol

[elifol]

Hi Seol,
I just posted a Q in the help section, and you seem to be the right person to answer, Im not sure how to best tag you in this conversation..

The question is what the cut-off value is based on, which method used. And what is the significance of Ir.eta, see picture

Thanks in advance for your time!
best,

[seol]

Hi elifol

With the following example,


The optimal cutoff value for Height can be found by substituting the value of lr.eta (0.545 ) into the above formula outcome, substituting the slope and intercept values for Height, and solving the equation.


Best
Seol

[seol]

Hi elifol

Compare the above results with Binomial Logistic Regression in jamovi below:

Best
Seol

[seol]

you will get the same results with the following analysis.

Best
Seol

[elifol]

Thank you!

I am more familiar with using Youden Index cut-off or closest topleft criterion, but this is another way of calculating the same thing? In my example the result is Yes or No (diagnosing a disease) and not a continous variable.

best,
Elin

[seol]

Hi

The main ROC analysis in snowCluster module is based on multipleROC R package.

For more information about multipleROC R package, see the following link: https://github.com/cardiomoon/multipleROC

Kind Regards
Seol