topic Re: proc pls : saving the model in Statistical Procedures

proc pls : saving the model

SASdevAnneMarie — Fri, 15 May 2026 17:09:32 GMT

Hello Experts,

When applying the PLS procedure, is there an option to save the trained model for application to the test set?

Thank you !

Re: proc pls : saving the model

PaigeMiller — Fri, 15 May 2026 18:27:05 GMT

@Rick_SAS gives the explanation of how to predict a test data set here: https://blogs.sas.com/content/iml/2014/02/17/the-missing-value-trick-for-scoring-a-regression-model.html

Re: proc pls : saving the model

SASdevAnneMarie — Fri, 15 May 2026 21:04:25 GMT

Thank you, but precisely, in Proc PLS we cannot use the train and test sets together in Proc.

Re: proc pls : saving the model

Ksharp — Sat, 16 May 2026 07:12:04 GMT

https://communities.sas.com/t5/Statistical-Procedures/score-for-PLS-procedure-issue/m-p/676606

The only thing you need to do is using SOLUTION option to save parameters estimator and score test data as PROC GLM .

Re: proc pls : saving the model

PaigeMiller — Sat, 16 May 2026 10:55:26 GMT

@Ksharp links to my exact solution for this problem. You would use the test data set in PROC SCORE, rather than the training data set, as I did in that example. PROC SCORE gives the predicted values using the PLS model. The final step where I create data set named COMPARE is not needed for the case where you want to score a test data set. The model is stored in data set COEFFS and COEFFS_T.

Re: proc pls : saving the model

PaigeMiller — Sat, 16 May 2026 10:59:38 GMT

@SASdevAnneMarie wrote:
Thank you, but precisely, in Proc PLS we cannot use the train and test sets together in Proc.

Yes, you can use the test and train data set together, in certain situations. Obviously you are not in that situation.

Re: proc pls : saving the model

SASdevAnneMarie — Sat, 16 May 2026 18:14:30 GMT

I found the answer to my question here:

Re: proc pls : saving the model

PaigeMiller — Sat, 16 May 2026 19:02:06 GMT

This is the same thing I linked to in my original post, which you said was wrong, you said you can't use the training data set and test data set in the same PROC PLS.