In article <[EMAIL PROTECTED]>,
Paige Miller <[EMAIL PROTECTED]> wrote:
> [EMAIL PROTECTED] wrote:
> >
> > I am attempting to build a performance attribution model and have
> > multicollinearity among my variables. As I would prefer not to
remove
> > variables, I am trying to program a PLS regression in SAS. I know
from
> > a previous posting that SAS has PROC PLS. MATLAB code to perform
PLS.
>
> You are trying to program PLS in SAS yet you also know SAS has PROC
PLS?
> Not clear to me what you mean. (It would have been clear to me if you
> had said "I am trying to use PROC PLS..." but that's not what you
said.)
>
Yes, I am using the PROC PLS function (correct diction?) to run partial
least squares regression in SAS. Please excuse my improper use of words
as I am just now learning to program in SAS.
> > First, I need to know if this code will perform the regression using
> > matrices. I believe Don Burrill posted this in a reply to a related
> > question.
>
> What matrices do you have? I believe that to perform PLS, you only
need
> a matrix X of the predictor variables and a matrix Y of the response
> variables. You don't need any other matrices to get started. What code
> are you referring to?
>
The reason I bring up multiple matrices is that I have read a paper
that talks about a method of solving for PLS by using steps of multiple
regressions. Basically, I understand the process except for the first
step which involves calculating a matrix of weights which maximize the
covariances between the predictor variables. Obviously, from your
response, the code must do all of the additional work. That is good. It
must be simpler to use than I first thought.
> > Second, do I need to understand the PLS process to be able to
program
> > it in SAS, or can I just understand the inputs/outputs and be ok?
(i.e.
> > do I need to understand the black box to be able use it?)
>
> If you are truly trying to write SAS code to program the PLS
algorithm,
> you probably need to understand the algorithm itself, as well as the
> inputs and outputs. However, it makes no sense to me for you to
program
> it yourself as PROC PLS exists and works. And if you are going to use
> PROC PLS (which isn't what you said), then you probably don't need to
> understand the algorithm. You do need to understand what PLS does, why
> it is the right technique (or a good technique) for your situation,
and
> you need to understand the outputs and be able to explain them to
> non-statisticians.
Yes, I agree. I do understand the what the process is doing, the inputs
and outputs, and why I should use it. What I didn't understand was how
to calculate the matrix that contains the variable weights that
maximize the covariance between the variables. Again, I do understand
the process, but I did not understand how to calculate all of its parts
(that is if I was doing this in steps). Since PROC PLS will do this in
one step, that shouldn't be a problem.
>
> If you view it as a "black box", which means to me that you throw in
> some data and get some results without understanding why, then you are
> probably headed for troube.
I am not do so blindly. I view it as a black box because it is doing
calculations that I can not do by myself.
Thank you for your responses.
>
> --
> Paige Miller
> Eastman Kodak Company
> [EMAIL PROTECTED]
>
> "It's nothing until I call it!" -- Bill Klem, NL Umpire
> "Those black-eyed peas tasted all right to me" -- Dixie Chicks
>
Sent via Deja.com http://www.deja.com/
Before you buy.
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
