Message-ID: <1287753176.2022.7.camel@ubuntu.ubuntu-domain>
Date: 2010-10-22T13:12:56Z
From: valerio.bartolino
Subject: resolution as random effect
Dear list,
I've a question about a spatial regression model I'm working with. Here
is my problem and question:
I'm modeling the local densities of a species (y) as a second order
polynom of the geographical position (lon,lat), time (year), and a
certain number of other variables:
mod1 <- lm( y ~ ... + poly(lon,lat, degree=2)*as.factor(year))
Then, I know that the sampling has a certain spatial resolution, that I
can represent with a regular grid. Thus, I was thinking that
observations within the same cell of the grid (gr.id) cannot be
considered completely independent, and that this could be represented
with a random effect like:
mod2 <- lme( y ~ ... + poly(lon,lat, degree=2)*as.factor(year),
random=list(gr.id=~1))
An additional level of complexity within my formulation, is given by the
temporal resolution. Thus, I would like to group the spatial dependency
of observations from the same cell within certain time intervals
(time.gr). Here my problem comes. I'm a bit confused on how to formulate
exactly the random effect expressed above into time.gr groups. I have
alternative slightly different formulations in mind, but I miss their
difference in practice.
Thanks in advance for any help or advice.
Valerio
--
Valerio Bartolino, PhD
Institute of Marine Research - Swedish Board of Fisheries
PO Box 4, 45321 Lysekil, Sweden
Department of Earth Sciences - Gothenburg University
PO Box 460, 40530 G?teborg, Sweden
e-mail: valerio.bartolino at gvc.gu.se
valerio.bartolino at gmail.com
><(((*> ----- ><(((*> ----- ><(((*> ----- ><(((*>