[R-sig-eco] ADONIS and pseudoreplications
Vít Syrovátka
syrovat at sci.muni.cz
Mon Nov 8 23:41:01 CET 2010
Hi listers,
I am facing a problem with pseudoreplications, similar to that by
Eduard Szöcs.
However, I want to try to explain the variation in species composition
by some environmental variables.
I decided to use adonis() for this, as there are many zeros in the
data-set, but don't know if it is correct, or how to do it so that it
was correct.
The design is:
Aquatic insect larvae were collected at 17 sites three times (spring,
summer, autumn), giving a total of 51 samples.
During each sampling campaign several chemical and substratum
variables were measured, while others were measured only once, as they
characterize the whole site and are not expected to change in time.
Now I want to relate the species composition to environmental
variables. I see that the samples are not independent, so I thought
about setting strata in the adonis model, but it didn't make me much
sense. I am interested in the differences among sites, therefore
strata should probably be season (?). Moreover, the effect of
environmental variables might differ among seasons, therefore
interaction with season should be probably included.
I ended up with this model:
adonis(log(spe+1) ~ env1 %in% season + env2 %in% season + ...,
distance= 'bray', permut= 999, data= env)
(where season is a factor with 3 levels)
I don't feel good about this model and would greatly appreciate any
suggestions on how to build an appropriate model.
Is there a way to analyze such data without splitting them into 3 data-
sets according to the season and analyze them separately?
Thanks in advance for any suggestions,
Vit Syrovatka
More information about the R-sig-ecology
mailing list