R-users,
Background: I took roadside samples of avian species richness and I
would like to model the relationship between species richness and
habitat around my 500 + sample locations (in a file called ROADSIDE).
However, one criticism is that roadside habitats do not represent
habitats throughout the study area. I tend to disagree because I detect
birds away from the roads and the pixel size is relatively course (30 x
30 m). I took 100 random samples (in a file called RANDOM) throughout
the study area and I would like to compare the different proportions of
each land cover type (variables are called GRASS, URBAN, PINE, HARDWOOD,
etc) between my sample and random sites. The distribution of
observations is highly skewed even after arc-sine-square root
transformations. The Wilcoxon test seems to be appropriate here but I
think I should use the bootstrap instead of just using 100 random
roadside samples. I did a search of R-help archives but nothing seemed
appropriate.
My question: How do I take 1000 bootstrap samples of 30 observations
from each group and use to Wilcoxon test to see if random samples differ
from roadside samples?
I have Efron and Tibshirani in hand but I do not have the MASS book
(coming in but not for weeks). I'm using a Windows XP on a MacBook and R
V 2.6.0.
Many thanks,
Jeff
[[alternative HTML version deleted]]
