I am studying factors that affect abundance and richness of waterbirds in brazilian lakes. I chose 12 lakes where I counted birds and collected environmental variables every month. all variables are numeric.
To fit models I'm using Generalized Linear Mixed Models with the "lme4" R package and selecting 90% confidence set through "MuMin".
My questions are:
(1) Are my data crossed or nested?
(2) Which would be the correct random term in the formula? Bellow, I provide some examples of alternative formula.
a
It looks like that the data structure is defined by two factors - lakes which define the 'spatial' dimension, and months which define the 'temporal' dimension. It is within this spatio-temporal data structure that you have collected data on environmental variables. If this is correct, then your data appears to have a crossed structure: Lakes x Months.
I don't think month is a random term as there is likely a seasonal effect (?). Also I am not too sure lakes should be treated as random considering they may not represent replicates. Without knowing your sampling regime, could there be differences with size\shore length, vegetation cover\type, lake depth, water quality etc?
Thank you for the answers! The lakes are my sampling units but they are not true replicates because, as almost everything in nature, they vary in several characteristics. Ok Stephen I will consult a statistician. Regards!
Hey there. Did you found an answer?
Probably I will use nested random factors but i'm not sure how I use. I know a priori that the size of effect and even the direction of effect vary among ecoregions and among cell, with cell, within ecoregion.
Since its a richness assessment, I intent to model it like this in lme4:
y ~ x + (x | cell : Ecoregion)
however I don't know if should I model like
y ~ x + (1 | cell : Ecoregion)
or even
y ~ x + (1 | cell : Ecoregion).
from the models you mentioned, which one have you used?
Hi Julian,
At that time, I used the glmmADMB package because I needed to apply a negative binomial distribution. This package has a somewhat different way to write the script.
You may find information about glmmADMB at the links below:
http://glmmadmb.r-forge.r-project.org/
http://glmmadmb.r-forge.r-project.org/glmmADMB.pdf
- Badges
- Science topic