Skip to content

About data distribution model

1 message · Nicholas Lewin-Koh

#
-----------------------------------------------------------------
Hi,
If your counts are relatively high, you might start with a log-normal or
gamma
distribution. What you are talking about here are species abundance
distributions
on which there is a large body of confusing (and often wrong)
literature.
A good resource is Steinhard Engen (1978), Stochastic Abundance Models,
CRC press.
But this may be hard to find these days, and is not for the beginner. 

But anyway, to just do an ANOVA, comparing treatments just taking the
log
and assuming a log-normal distribution may be good enough. By the way
what is your response? Number of species? In that case you could do

fit<-lm(log(No.Species)~Habitat, mydat)
anova(fit)
or
glm(No.Species~Habitat, mydat,family=poisson())
anova(fit)

But I am not sure this answers anything. If your habitats
follow a gradient, or you have covariates describing the habitats, you
may want to look at ordination methods in vegan or ade4

good luck

Nicholas