Dear members,
I would need expertise and advices about the
multicolinearity problem.
I link a dependent variable with five regressors or
independent variables. I use multiple linear regression in order to examine the
relative contribution of the regressors. Unfortunatly, the regressors are
correlated and I know that in such case, the individual coefficients can not be
interpreted as the relative contibution of the variables.
My questions are:
Is the regression on principle component a solution? I
mean, can I draw any conclusions on the relative importance of two particular
regressors by examining either their contribution on the principle component or
by examining their coefficients with respect to the linear regression performed
on the principle components?
Is there any other solution ?
Thanks for your help
Chris
============================
Please note my new Email
adress
============================
---------------------------------------------------------
Dr
Nguyen Christophe
Unité associée Agronomie &
Environnement
INRA-ENSAIA
BP 172
F54505 Vandoeuvre les
Nancy
FRANCE
Phone : (33) 383 59 57 87
Fax : (33) 383 59 57
99
Email : [log in to unmask]
---------------------------------------------------------