Abraham,
On Mon, 1 Oct 2007 11:48:01 -0700,
"Abraham de Alba A." <[EMAIL PROTECTED]> wrote:
> Dear Ecologgers: I have been doing some regressions with temperature
> means across time series on about 80 stations in a rather small region
> (less than 2 degrees lat), contrasting before and after 1991, and not
> surprisingly I have found quite significant slopes in many stations. But
> in trying to make out a pattern in statistical terms I am not sure if I
> am breaking some rules, for one I don't know what to do with stations
> where there are NO significant trends, they are not zeros. So I can
> think of two approaches, multiple regression using altitude, lat, long,
> even population as independant variables or use a geostatistical model
> in ArcView or any other geostatistical software.
This is an appropriate problem for mixed effects models. You'll have to
check whether the trends should be considered linear or non-linear, but
the best reference on the subject is Pinheiro and Bates book:
@BOOK{1914,
title = {Mixed-effects models in S and S-PLUS},
publisher = {Springer-Verlag},
year = {2000},
author = {Pinheiro, J. C. and Bates, D. M.},
address = {New York, NY},
}
Best,
--
Sebastian P. Luque
