topic Is there any way to get the list of variables used for each fold for cross-validation? in SAS Data Science

Is there any way to get the list of variables used for each fold for cross-validation?

mzhao — Fri, 15 Nov 2019 14:11:09 GMT

Hi SAS Community,

I have performed cross-validation on a logistic regression model in SAS Enterprise Miner and I found the logs did not report the variables that were used in each cross-validation iteration. Is there any way of obtaining the information on the variables used?

The process flow goes like this: ... -> Transform Node (Formula for cross-validation) -> start groups -> Regression Node -> end groups -> model import -> ...

Thanks!

Re: Is there any way to get the list of variables used for each fold for cross-validation?

Reeza — Thu, 14 Nov 2019 22:45:45 GMT

I don't know exactly what you're doing, but typically cross validations doesn't involve changing variables but changing observations.

Re: Is there any way to get the list of variables used for each fold for cross-validation?

mzhao — Fri, 15 Nov 2019 13:56:14 GMT

I don't want to change the variables or observations here. I want to see if there's a way to view the variables that the regression model has used in each cross-validation iteration.

Re: Is there any way to get the list of variables used for each fold for cross-validation?

Reeza — Fri, 15 Nov 2019 16:22:30 GMT

AFAIK, cross validation doesn't change the variables in each iteration, it changes the observations. It's a resampling procedure unless you're doing something drastically different.

Re: Is there any way to get the list of variables used for each fold for cross-validation?

mzhao — Fri, 15 Nov 2019 16:26:07 GMT

Since my logistic regression model is using stepwise regression, the variables used will change if the sample used to train the data changes. So, if the training sample changes, the variables will change in this case.

Re: Is there any way to get the list of variables used for each fold for cross-validation?

Reeza — Fri, 15 Nov 2019 16:29:35 GMT

You never mentioned step-wise regression anywhere so there's no way I can know that was selected. Look at the options for the logistic regression, there should be one to show more details for each iteration, since that exists with the proc.