[R] nested mixed-effect model: variance components

[Eric Pante]

Dear listers,

I am trying to assess variance components for a nested, mixed-effects 
model. I think I got an answer that make sense from R, but I have a 
warning message and I wanted to check that what I am looking at is 
actually what I need:

my data are organized as transects within stations, stations within 
habitats, habitats within lagoons.
lagoons: random, habitats: fixed
the question is: how much variation is due to lagoons? habitats? 
lagoons*habitat? transects?

Here is my code:

res <- aov(COVER ~ HABITAT + Error(HABITAT+LAGOON+LAGOON/HABITAT), 
data=cov)
summary(res)

and I get Sum Sq for each to calculate variance components:

Error: STRATE
        Df Sum Sq Mean Sq
STRATE  5 4493.1   898.6

Error: ATOLL
           Df Sum Sq Mean Sq F value Pr(>F)
Residuals  5 3340.5   668.1

Error: STRATE:ATOLL
           Df  Sum Sq Mean Sq F value Pr(>F)
Residuals 18 2442.71  135.71

Error: Within
            Df Sum Sq Mean Sq F value Pr(>F)
Residuals 145 6422.0    44.3

My error message seems to come from the LAGOON/HABITAT, the Error is 
computed.
Warning message: Error() model is singular in: aov(COVER ~ HABITAT + 
Error(HABITAT+LAGOON+LAGOON/HABITAT), data=cov),

THANKS !!!
eric



Eric Pante
----------------------------------------------------------------
College of Charleston, Grice Marine Laboratory
205 Fort Johnson Road, Charleston SC 29412
Phone: 843-953-9190 (lab)  -9200 (main office)
----------------------------------------------------------------

	"On ne force pas la curiosite, on l'eveille ..."
	Daniel Pennac