0
$\begingroup$

experts, I have a question about PLS procedure in SAS. The manual said that the prediction on new data is by combining training data and new data (new data dont have response values).

I did experiments, say, 300 training data and 2000 new data, by using that approach, the predicted results of 2000 new data are quit similar (variability is less than 0.3) which is not right. Can you have some suggestions on how to do prediction using PLS procedure in a proper way?

Many thanks for your kind reply.

$\endgroup$
2
$\begingroup$
  • The similarity of the output in itself does not allow any conclusions. After all, that may be correct.
  • PLS regularization "pushes" the output towards the mean output in the training data. A PLS model that too heavily regularized (undercomplex) will typically predict with a bias towards the mean, i.e. too low variability. Undercomplex here is wrt. the underlying truth. The undercomplex model can still be the objectively best model that can be obtained with the given training data set.
$\endgroup$

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.