From: merguez99@yahoo.fr   
      
   On Jul 16, 12:16 pm, Ted Dunning  wrote:   
   > On Jul 14, 9:29 pm, Francois  wrote:   
   >   
   > > I am looking for a regression model that can predict multiple   
   > > variables at the same time, i.e. multiple dependent variables.   
   >   
   > Take a look at latent variable models.   
   >   
   > Also, can't this problem just be decomposed into a conventional   
   > regression on a single dependent variable?  For a linear problem, you   
   > seem to be wanting to find a minimum error solution of this system:   
   >   
   > A_1 x = b_1   
   > A_2 x = b_2   
   > ...   
   > A_n x = b_n   
   >   
   > This could be stacked into a single system using:   
   >   
   > [A_1 ; A_2 ; ... ; A_n] x = [b_1; b_2; ... ; b_n]   
   >   
   > (where ; indicates vertical stacking as in matlab or octave).   
   >   
   > Other kinds of regression typically involve a link function of some   
   > kind and possibly multiple levels of linear combination and linking.   
   > Regardless of that, this stacking trick should still work just fine.   
   >   
   > So doesn't this mean that very ordinary techniques will work?   
   >   
      
   Thanks a lot, I did indeed try with simple linear system resolution   
   (using SVD), however the linear model doesn't seem to fit that data as   
   well as some non-linear models, so I was wondering if there existed   
   non-linear regression models that support multiple output variables.   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)