topic obtaining the principal components from proc princomp in Statistical Procedures

obtaining the principal components from proc princomp

ilikesas — Thu, 09 Mar 2017 03:06:52 GMT

Hi,

I have been using proc princomp to perform a principal components analysis. It gave me the eigen vectors, but not the principal components of the original data. Is it possible to transform the original data in order to get the princpal components and later use them in a regression?

Thank you

Re: obtaining the principal components from proc princomp

Reeza — Thu, 09 Mar 2017 04:01:18 GMT

What does your code look like?

Can we assume you've tried the standard output datasets and they don't have the values?

Re: obtaining the principal components from proc princomp

Ksharp — Thu, 09 Mar 2017 04:28:08 GMT

Check this out.

http://blogs.sas.com/content/iml/2014/11/07/distribution-of-blood-types.html

Re: obtaining the principal components from proc princomp

Rick_SAS — Thu, 09 Mar 2017 10:42:03 GMT

If you look at the Getting Started example in the PROC PRINCOMP documentation, you will see that

the eigenvectors are listed in the "Eigenvectors" table. They are named Prin1, Prin2, Prin3, etc.

You can use the OUTSTAT= option on the PROC PRINCOMP statement to create an output data set that contains the eigenvectors.

They are listed as the rows that are labeled PRIN1, PRIN2, etc.

If you want to use the PCs in a regression, I believe you want to use the SCORES, which are available in the OUT= data set.

Re: obtaining the principal components from proc princomp

ilikesas — Fri, 10 Mar 2017 03:32:32 GMT

Thanks Rick, when I added "out = new_data" I obtained a data table which contains the original variables and the newly calculated principal components.

Could I now use these Principal Components in a regression? I have a IML code which does the Principal Components Analysis, but for some reason standardizes the Principal Components - but the PC's that I obtained above are not standardized, so I just want to make sure that these non standardized PCs are the ones that I should use in a regression.

Thanks