Dears UseRs,
In Zuur's book (Mixed effects models and extensions in ecology with R), p
167, it is mentionned that the Variogram function (package nlme) scale the
sill to 1.
Here is the plot of the semi-variogram given by the Variogram function on my
"lme" object. (http://imm.io/3OLe)
Note that the semi-variance exceed 1 !!
It can probably explain why the Gaussian correlation structure fitted in my
"lme" model is strange (curve is probably fitted to the data above 1).
See my previous message a few minutes ago untitled "Variogram (nlme) of a
lme object - corSpatial element question."
I don't know if it is a bug, but Implications could be important for model
fitted with a spatial correlation structure.
Arnaud
[[alternative HTML version deleted]]
