Date: Sat, 25 Sep 2010 18:49:13 -0700
Reply-To: "David J. Moriarty" <djmoriarty@CSUPOMONA.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: "David J. Moriarty" <djmoriarty@CSUPOMONA.EDU>
Subject: Re: Proc PRINCOMP
In-Reply-To: <AANLkTi=mxVTBjJ9Ke_AL8jYaF4z-yfuB5DAEqQhZF+n-@mail.gmail.c om>
Content-Type: text/plain; charset="us-ascii"; format=flowed
At 05:55 AM Saturday 9/25/2010, Michael Bryce Herrington wrote:
>Is there an option with PROC PRINCOMP that would allow me to score the
>validation data set with the formula created by running the procedure on the
>training data? Unfortunately, this is my first time working with principle
>component and I do not really know what is happening. Any info would be
>appreciated.
I don't know of an option that allows this. The statisticians on the
list will no doubt have better suggestions, but for your consideration:
You might want to see if the relationships between the variables and
components are the same for the training and validation data sets.
Perhaps you could consider doing something like a Mantel test between
the eigenvector matrices (variables x eigenvectors) for the two data
sets. If you prefer, you could use PROC FACTOR which would produce
correlations between the components and variables (the "Factor
Pattern" matrix). The information is the same, but some are more
comfortable with the interpretation of the correlation coefficient
than the elements of the eigenvector matrix. If you use PROC FACTOR,
you'll probably need to use the N= option to get output on all the
components, my recollection is that the default in FACTOR is a
stopping rule of 1.0 for the minimum eigenvalue to be included in the output.
Comparing the eigenvalues for the two data sets might also be of
interest. I'm not sure what might be a recommended method here. My
first inclination would to do some sort of paired analysis, using
randomization to produce a p value - if you want a p value.
HTH
Dave
-------------------------------------------
David J. Moriarty, Ph.D.
Professor and Graduate Coordinator
Biological Sciences Department
California State Polytechnic University
Pomona, CA 91768-2557 USA
http://www.csupomona.edu/~djmoriarty/
