[S] new to spatial stats

Geoffrey Phelps (gphelps@best.com)
Fri, 23 Oct 1998 20:32:47 -0700


fellow S users,
Being new to spatial statistics, I am looking for references that explain
spatial statistics
at the conceptual level rather than the mathematical level. The following
are examples of
the types of questions that I would like to be able to answer:
- given a scattered point dataset in 2D space, under what types of
circumstances
would a spline(penalized least squares vs. regression splines,
etc.)/loess/trend
surface/kriged surface be most appropriate? And why (keeping in mind
that I
must justify my choice to collegues who know less about statistics than
I do)?
-given a "best fit" surface (for example a trend surface) what is the
appropriate
method of assessing the range of error for a particular point on the
surface?
-given two surfaces, how would one determine the degree of correlation in
a given
spatial window (ie: are two surfaces positively, negetively or not
correlated over
a given X-Y window)?
-What sort of sampling distribution and quantity is generally considered
reasonable
(I realise there is some debate on the subject) for estimating a
variogram? How
would one test the assumption?
-Given a scattered (or regular) dataset, what issues are involved in
estimating the
appropriate resolution of derivative data (e.g. given a scattered dataset
with an
average sampling interval of 200m, is an output resolution of 200m
justified?
What sort of sampling distributions would justify/not justify this? How
would
one test a sample for meeting this criterion?)
-Are there any examples of "statistics gone awry", where failing to take
spatial
relations into account resulted in misleading results (again keeping in
mind that
I must explain my reasoning to collegues, and it helps to be concrete).

I appreciate any help. If there are any references to S-plus code (esp.
spatial stats)
they are most welcome.

Thank you,
geoff phelps

gphelps@best.com
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news