Re: PLS2

From: Jean Thioulouse (Jean.Thioulouse@biomserv.univ-lyon1.fr)
Date: Thu Jan 21 1999 - 16:28:03 MET


Agustin Lobo <alobo@ija.csic.es> wrote:
>I'm reading the doc for PLSgen2 by Luc Monimeau.
>Both the explanatory (X) and dependent matrices are
>20r x 3cols, but then the tk explanatory components are 20 x 4,
>why do we have one more component than initial variables?

This is an error in the listing. It has of course only 3 columns.

>Instead, in the "fiche" by Chessel and Monimeau, X is
>39 x 3 and Y is 39 x 15 (a better choice for an example, I think)
>and the VI.ivl1 (which I understand is the LS2 components of X,
>the file naming is very confusing, why not just X.PL2 ?)
>is 39 x 2 (becasue they select to keep only the first 2 components,
>I guess).

This is not in the PLS regression part of the document I think,
but in the part about instrumental variables (hence the "iv").

>Finally, Monimeau uses PLSgen2:Modelling, while Chessel and Monimeau use
>PCA on instrumental variables. Is this just a difference on soft versions?

No, principal component analysis on intrumental variables and
PLS regression are two different things. PCAIV is a PCA after
projection on a subspace. This is the same family of methods
as CCA (canonical correspondence analysis). PLS regression is
rather like co-inertia analysis methods.

Jean

--
Jean Thioulouse - Laboratoire de Biometrie -  Universite Lyon 1
69622 Villeurbanne Cedex - France           Fax: 04 78 89 27 19
Tel: 04 72 43 29 01 http://pbil.univ-lyon1.fr/ADE-4/JTHome.html



This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:35:55 MET