I am using PROC PLS in SAS with multiple independent variables and multiple dependent variables. I would like to know how my independent variables are contributing to the scores for the first couple of generated components.

For example, if I was using principal components instead of partial least squares, I would look at the eigenvectors (principal component loadings) to determine this. Is there something similar I can look at for partial least squares? In the SAS documentation it refers to something called a "weight vector" that seems to be roughly what I want - but there does not appear to be any way of extracting it.

General suggestions for approaching this problem would also be appreciated!


1 Answer 1


The weight vectors from PROC PLS can be extracted using the "details" option.

However, they are not particularly interpretable because they relate the components to the "deflated" X matrix, which changes after each component is computed. I found that using the SIMPLS method (done by specifying method=SIMPLS) gives me exactly what I want - variation in the dependent variables is accounted for when constructing the components, but the components are a linear combination of the original X matrix, and thus there are interpretable loading vectors to look at.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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