I am trying to use mixed effects model with lme function in S-Plus 2000.
I will try to explain my sampling design:
I want to study the impact of afforestation on
the abundance of one bird species in adjacent
agricultural habitats. So, I selected 52 forest
patches with an adjacent fallow field
(agricultural habitat). In order to evaluate the
variation in abundances with increasing distance
from the forest edge, the bird counts were undertaken as follows:
1) for each SITE (forest + adjacent fallow) we
performed 4 point counts along a transect
radiating from the forest edge, with the first
one located at the forest border and the other
ones spaced at 100 m intervals (100, 200, 300m);
this was coded as a continuous variable (Dist)
2) I have three main types of forest (eucalyptus,
pinewoods and oakwoods), which were coded as two dummy variables (X1 and X2);
3) I also have landscape context variables (e.g.,
proportional cover by potentially suitable
habitat within a 1km buffer from the transect;
coded as a continuous variable A), which have the
same value for all distance classes within each of the 52 transects.
Again, I'm interested in understanding if bird
abundances changed with increasing distance from
the forest edge, and whether such variation is
influenced by forest type and landscape context.
I'm not sure how to specify such design.
Presently I'm using the following model, though I
don't know whether this is correct or not
model<-lme(abundance~Dist+X1+X2+A+X1xDist+X2xDist+AxDist,
random=~1 | Site, data=file)
My main question is whether Dist should also be
used as a random component, because all Dist
values within each site have the same values of X1, X2 and A.
best regards,
Luis Reino
-------------------------------------------------------
Luís Miguel Reino
Centro de Estudos Florestais
Departamento de Engenharia Florestal
Instituto Superior de Agronomia
Tapada da Ajuda
1349-017 Lisboa
Portugal
|