topic Re: Validation for Logistic regression model JMP 15 in Discussions

Validation for Logistic regression model JMP 15

AnastasiaMan — Sat, 10 Jun 2023 20:42:54 GMT

I saw many comments on how to run a validation for a jmp pro version, that unfortunately I don't have.

In jmp 15, when I open the Fit Model, there is no box to add the validation column.

So, what I did was to create manually a validation column, with 60% data for training, 20% for validation and 20% for testing.

I build the model based on the training dataset, and I can see the predictions for the validation and testing set.

How can I generate the R^2 and other fit details for the validation dataset on the model that is build on the training set?

Re: Validation for Logistic regression model JMP 15

dale_lehman — Mon, 01 Feb 2021 14:43:36 GMT

I am assuming you excluded the test data when fitting the model (I'm not sure what you did about the validation data - if you excluded that as well, then the test data is redundant - otherwise, I'm not sure what purpose the validation data served in your analysis). Unexclude all the rows and use Fit Y by X, with Y as the predictions and X as the actual values - and put Validation in the By box. You will get separate plots (and R-squared) for each subset.

Re: Validation for Logistic regression model JMP 15

AnastasiaMan — Tue, 02 Feb 2021 10:36:17 GMT

Hi and thank you, it does work.

But does it calculate 3 different models based on each subset?

What I initially did, was to use only the training set for building the model, and as I have more than one possible models, I want to fit them in the validation data to choose the best one. So, I need to apply the same model built on the training dataset, to the validation dataset and compare which fits the best.

Re: Validation for Logistic regression model JMP 15

Zhiwu — Wed, 03 Feb 2021 10:14:45 GMT

You can first fit your model using training dataset, the save prediction formula for each of models. Then using fit Y by X with your Y value in validation set and test set and X value is prediction formula from you models. Then by doing fit line or fit special with 45 degree line you will get all statistics number in output.