Dear R users,
I have a question regarding accounting for spatioautocorrelation in the
residuals of gls models. I would like the error term of my GLS model (in
the nlme package) to account/control for spatialautocorrelation among my
sample units. So I use a model that looks something along the lines of:
## specify correlation structure
Cor1<-corSpher(form = ~Lat + Lon, nugget = TRUE)
#### model
m1<-gls(InvSpCt ~ lPD1900 + lPD2010 + lPC2012 + DP + lRd, method = "ML",
correlation = Cor1)
What the response and explanatory variables are don't matter for the
purpose of my inquiry. My sample size is very large (N ~2500). For this
reason, it is simply not possible to run this model (at least not from my
computer!). What would make it possible is if I could limit the distance
at which spatioautocorrlation is accounted for. For example, if I could
account for spatioautocorrlation up to a certain distance, but disregard it
after that point. Inspection of variograms justifies this approach.
So my question....is there a way to limit the distance at which the spatial
correlation function is applied to the error term in gls models. I read
the R documentation and have searched online but have found no mention of
such a 'creature'. NOTE: I am also trying to work with SARerr models, but
I thought I would look at the data using a statistical modeling approach
that I am more familiar with as well.
Thank you in advance for your help. Also, thank you in advance for your
patience if I am missing some major point here in how gls models
incorporate spatioautocorrelation in the error term.
--
Basil V. Iannone III, Ph.D.
Department of Forestry and Natural Resources
Purdue University
195 Marsteller St.
West Lafayette, IN 47907
765 - 543 - 2081
[[alternative HTML version deleted]]
______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
