Proc PLS Response variable

[Daisy2]

For PLS does my Response (Y) need to be Normal?  I’m wondering whether I need to use my transformed Y or is it OK to use the measured data with no transformation?  Thanks.

[Ksharp · Accepted solution]

No need to be normal and no need transformation .

PLS is totally different algorithm with Least Square algorithm .

[PaigeMiller]

But on the other hand, the Y variable cannot be binary or categorical, PLS doesn't work well that way (although I have done it).

[Daisy2]

Thanks, I see that my X's need to be centered and scaled - is that right? If so, how do I do that?

[Ksharp]

Yes. It is true. SAS will do it for you . you don't need to do anything.

" be centered and scaled" is a part of PLS algorithm .

[PaigeMiller]

---

@Daisy2 wrote:
Thanks, I see that my X's need to be centered and scaled - is that right? If so, how do I do that?

---

They don't "need to be" scaled, but that's usually the way PLS is run. However, there are valid reasons why you might not want to scale the x-variables in some scenarios. So PROC PLS gives you the choice of using scaling, which is the default, or not scaling, which is the NOSCALE option.