Dear all
I would like to fit a linear model on a response variable Y =
(Y_1,..,Y_n) against a particular continuous variable. The values of the
response variable are the estimated coefficients of a factor
(Y_1=Beta_1,..Y_n=Beta_n) obtained with a previous GLM. Given the
contrasts (treatment), these estimations of Beta_i are correlated. To
take account of these correlations, I decide to use the function gls
with a correlation matrix. Does it has any sense?
I have extracted from the covariance matrix of the first GLM, the
covariance between the Beta_i and I try to find (without success!) a way
to introduce this matrix in the correlation structure of the gls
function.
Can anybody help me?
Any solutions will be appreciated and sorry for the bad english.
thank you very much.
(S+2000)
--
......................................................................
Stephanie MAHEVAS (Stephanie.Mahevas@ifremer.fr)
IFREMER/MAERHA Tel: 02 40 37 41 81 Fax: 02 40 37 40 75
(Mathématiques Appliquées à l'Evaluation
des Ressources Halieutiques et Aquacoles)
rue de l'île d'Yeu
BP 21105 44311 NANTES Cedex 03
http://www.ifremer.fr/maerha
o \ o / _ o __| \ / |__ o _ \ o / o
/|\ | /\ ___\o \o | o/ o/__ /\ | /|\
/ \ / \ | \ /) | ( \ /o\ / ) | (\ / | / \ / \
......................................................................
|