[R] gamm (package mgcv) with large datasets
jmburgos at u.washington.edu
Wed Jul 18 21:41:15 CEST 2007
I am interested in fitting a Generalized Additive Mixed Model with
spatially correlated errors to a large, spatially indexed, data set
My initial analysis was a Generalized Additive Model that included a two
dimensional smooth term to model spatially correlated effect (i.e.
s(latitude,longitude)). The problem is that the residuals of this model
are still spatially correlated, so it seems that I should use a GAMM in
which the spatial autocorrelation is modeled explicitly.
The problem is that, as stated in the documentation of the mgcv package,
my dataset is too large for the gamm function. Is anybody aware of an
alternative approach to analyze this data?
Julian M. Burgos
Fisheries Acoustics Research Lab
School of Aquatic and Fishery Science
University of Washington
1122 NE Boat Street
Seattle, WA 98105
More information about the R-help