proc pls: center and scale

[SASdevAnneMarie]

Hello Experts,

I am wondering how to center and scale the X variables and center the Y variables for the proc pls data=train method=pcr application. Are there specific options in SAS for this?

Thank you!

[PaigeMiller]

I believe you want the VARSCALE option in PROC PLS.

VARSCALE

specifies that continuous model variables be centered and scaled prior to centering and scaling the model effects in which they are involved. The rescaling specified by the VARSCALE option is sometimes more appropriate if the model involves crossproducts between model variables; however, the VARSCALE option still might not produce the model you expect. For more information, see the section Centering and Scaling.

For more advanced centering and scaling methods, use PROC STDIZE.

I always feel uncomfortable when PCR is chosen for METHOD, as this logically (based upon the criterion being optimized) will produce worse fits than METHOD=PLS. Maybe there are valid reasons for using PCR instead of PLS, but in general PCR will not fit as well as PLS.

[SASdevAnneMarie]

Thank you Paige Miller!

I thought that proc pls centered and scaled variables by default.

This was precisely to compare the PCR and PLS outputs.

[PaigeMiller · Accepted solution]

---

@SASdevAnneMarie wrote:

I thought that proc pls centered and scaled variables by default.

---

Yes it does. My mistake, VARSCALE is an additional option to help with crossproducts of model variables (if you have any).

This was precisely to compare the PCR and PLS outputs.

Outputs as in measures of fit? Because there's really no point in comparing loadings or scores between PCR and PLS.

[SASdevAnneMarie]

I would like to compare the pcr et pls methods for academic purposes.