Re: PLS1-PLS2-PCAVI-MLR : comment choisir ? (fwd)

From: Jean Thioulouse (Jean.Thioulouse@biomserv.univ-lyon1.fr)
Date: Tue Jan 26 1999 - 17:42:06 MET


Agustin Lobo <alobo@ija.csic.es> wrote:
>Et en plus on a Canonical Correlation, in which both X and Y are transformed
>into X.can Y.can to maximize correlation between pairs of variables.
>Where would this method fit in your scheme?

Canonical correlation analysis is, like co-inertia analysis, a symetric method.
I.e., both tables play the same role : there is not a set of dependent variables
and a set of explanatory/independent ones.

>Also, should not prediction done through CaCorr
>work better than PLS2 as Cancorr transforms both sets?

No. PLS2 is mostly usefull when you have many independent variables. You
cannot use Cancor in this case.

--
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