Thank you all,

I think I am a little bit confused and I am not sure I really know what is appropriate for what I need. I have some data with a lot of missings where the independent variables are highly correlated and as far as I understand the PLS regression (partial least squares - if I get it right) would be more fitted than multivariate regression or GLM, because of the multicolinearity. I hope it cleares a little what my problem is, because for me....

Thank you again.

Vlad.

On 6/29/06, Dan Zetu <dzetu@hotmail.com> wrote: > > First, I am not aware of any SPSS capability to do PLS regression. SAS has > proc pls that does PLS regression, not to be confused with the PLS > approach > for latent variable modeling. If you have SAS, then its documentation has > a > rather good description of the procedure and what it does. > > Although I personally haven't seen one, any chemometrics book should > contain > at least some brief description of the PLS regression procedure, as it is > this field in which PLS has found many applications. Also, SAS researchers > have published papers in the past on PLS regression (they should be > publicly > available on SAS website). > > Hope this helps, > > Dan > > > >From: vlad simion <vlad.simion@gmail.com> > >Reply-To: vlad simion <vlad.simion@gmail.com> > >To: SPSSX-L@LISTSERV.UGA.EDU > >Subject: pls regression > >Date: Thu, 29 Jun 2006 09:15:49 +0300 > > > >Hi all, > > > >could anyone help me with some clearings on the PLS Regression, the > >algorithm, what are the assumptions for running it and so on. And where > can > >I find some materials on pls regression? > > > >Thank you in advance. > > > >Vlad. > > > >-- > >Vlad Simion > >Data Analyst > >Tel: +40 720130611 > > >

-- Vlad Simion Data Analyst Tel: +40 720130611