0
$\begingroup$

I am interested in running a partial least squares analysis using PROC PLS in SAS 9.4. I understand that, by default, the predictors and response variables in PLS are centered to a mean 0 and scaled to a standard deviation of 1. I have read that the PLS procedure can accommodate categorical predictors, and their is a CLASS statement available in PROC PLS.

My question is how the categorical predictors are handled in this procedure. Are they also centered and scaled by default? Is that appropriate or do I need to ensure they are not centered and scaled. Thanks!

$\endgroup$
1
$\begingroup$

I am not a SAS user but the general practice I have encountered so far indicates that since PLSR and PLS-DA (PLSR where response variables are categorical and one-hot encoded) tries to maximize covariance between predictors and responses, mean centering is mathematically necessary. Scaling is, as alwasy, optional.

However, there is one article discussing whether the centering is necessary even though it does not specifically mention about categorical variables. See this link

I believe categorical variables are handled as if they were continous variables.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.