UNASMITH@pucc.Princeton.EDU (Una Smith) (05/18/89)
I want to study the spatial distribution patterns of individuals
(trees) in a 2D space (an area of forest). There are a number of
popular methods for characterizing clumping, scatter, etc., but I
know very little about the relative merits of each method. I hope
to generate some useful discussions about measures of spatial
distribution (in any context) and to learn something new that may
apply to my own work.
I would appreciate current references to published studies,
and any ideas about interesting dynamics to look for.
Methods:
One popular method seems to be to calculate a statistical index
of the deviation of local density from an expected value (where
the expected follows a binomial or other distribution). This
method is less computationally expensive than some others, but I
have serious questions about its descriptive power. An advantage
is its relative independence from edge effects.
Another method is to calculate pairwise distances between every
individual and every other individual, or to class individuals
into cells, and treat the cells in a pairwise fashion. This
method can be very expensive, since for N individuals, there are
N(N+1)/2 pairs. Also, the frequency of pairs at a given distance
is highly affected by the relative size (and shape) of the sample
area. I understand that packages such as NAG and ESSL provide 2D
Fourier transform subroutines that would be appropriate to my
study, and that each has a built-in (?) method of adjusting for
edge effects. I believe these involve wrap-around measures at
sample edges. Another method, used often in astrophysics, is a
statistical function which assigns relative weights to each
distance to correct for edge effects. Would anyone care to
comment on these methods and their relative merits?
Are there other techniques that I haven't mentioned? Please
comment! Reply directly to me, and indicate where you read this
note, so I'll have an idea of where to post any summaries.
This note is being posted to bionet.population-bio
comp.lang.fortran
sci.astro
sci.math.stat
sci.physics
Una Smith UNASMITH@PUCC.BITNET or unasmith@pucc.princeton.edu
Program in Ecology, Evolution and Behavior
Department of Biology, Princeton University