Programming the statistical procedures from SAS

Help: external cross validation with proc glmselect

Reply
User
Posts: 1

Help: external cross validation with proc glmselect

Hi everyone. I am trying to predict an outcome with several predictor variables (4-5), and some of them show a certain level of collinearity (up to an r correlation coefficient of 0.7) so I must apply a restricted regression technique. I am using elastic net because I understang it should have an advantage over lasso or ridge in this situation.

Well, with proc glmselect I am able to run an elastic net regression with external k fold cross validation as the model selection method

http://support.sas.com/documentation/cdl/en/statug/67523/HTML/default/viewer.htm#statug_glmselect_de...

 

Now, I would like to graph predicted vs observed values (and then calculate slope and intercept for the predicted vs observed regression) for all the data used in the k fold steps, but I cannot obtain the parameters estimates for each k fold run.

The cvdetails option in the model statement indeed works for obtaining parameters estimates when running internal cross validation, but it does not for external cross validation. Any clue?

 

Thanks in advance,

Pedro

Occasional Learner
Posts: 1

Re: Help: external cross validation with proc glmselect

Hi.

 

I've right now exact the same question as Pedro in 2015. At that time no answer was posted.  Does anyone have an idea by now? I'm interested in the predictors selected by 5 fold corss validation in each of the 5 steps by proc glmselect...cvdeatils doesn't work with selection=elasticnet(choose=cvex) .



Hoping for any advice.

Ask a Question
Discussion stats
  • 1 reply
  • 237 views
  • 0 likes
  • 2 in conversation