Dear All,
I am using S-plus 2000 under windows XP.
I am using nlme to fit growth functions to tree growth data. As I need
estimates for standard-errors of random-effects, I tried with
"object$apVar" (object being of class nlme) supposed to give "an
approximate covariance matrix for the variance-covariance coefficients".
The estimates I obtained were by far to high.
In order to check the meaning of these so called variances for
variance-covariance coefficients, I adjusted the same model, rescaling
parameters (for instance p' = p/100). The new estimate of the mean of p' is
100 times the old one, as well as for the random effect (sigma p'). But
using $apVar, I found exactly the same estimates for the precision of sigma
p' !
Can anyone please tell me how is $apVar parametrized ? What is exactly
given inside ?
With great many thanks,
JD Bontemps
--
Jean-Daniel BONTEMPS
UMR 1092 Inra-Engref "LERFoB" - équipe Ecologie Forestière
ENGREF - 14, rue Girardet
F-54000 NANCY
tél: (+33) (0)3 83 39 68 75
mél: bontemps@engref.fr
www.nancy.inra.fr/lerfob/
|