Dear Mixed-Modelers,
Sorry to re-post this message.
I am trying to fit linear models using lme and considering the spatial
autocorrelation structure of my data.
I know that the correlation has a half-normal shape, so I use the function
corGaus.
Observations are coming from several surveys, done at various dates
(sometimes several surveys occurred at one date), done during several years
(sometimes several surveys by year).
- I want to consider the spatial correlation that occur between observations
done within EACH survey
- I DO NOT want to consider spatial correlation between observations done
during ALL surveys
I define the spatial correlation structure as: corGaus(form = ~ X + Y |
Year/Date/Survey)
Is that correct ?
Another point ... outside of the function lme, I can define the range
applied in the corGaus function as:
corGaus( range value, form = ~ X + Y | Year/Date/Survey)
But when I include this statement in the lme function, it do not take into
account the range value given and estimate its value.
Is there a way to force the range applied on the corGaus structure used in
lme in order to apply exactly the same autocorrelation structure to all my
models.
Thanks for your help !
Arnaud
[[alternative HTML version deleted]]
_______________________________________________
R-sig-Geo mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
