[R-sig-eco] Spatial Autocorrelation - using co-ordinates as smooth factors in GAM's plus Spatial correlograms + Moran's I
Edzer Pebesma
edzer.pebesma at uni-muenster.de
Thu Jan 27 22:14:13 CET 2011
Dear Chris,
to me, it is never a questions whether spatial correlation is present
(it is), but whether it is strong enough to address it. I never
understood why people would test for spatial autocorrelation, as
non-significant autocorrelation will only indicate that the sample size
is small. Of course, this is with my geostatistical glasses on, through
which I look at the world.
Including coordinates in a GAM may carry the risk that the solution is
not rotation invariant: directions of x and y are usually arbitrary
(unless on global scale), and would you be happy if the two 45 degrees
coordinates (x+y) and (x-y) would result in a different model? It seems
two-dimensional smoothing splines in x and y work well, in that respect;
I believe gam in package mgcv has these. This would filter out the
low-frequency, smooth trends; variogram modelling and kriging of
residuals may be better at picking up more local deviations.
When lost in the kriging woods, there's always r-sig-geo.
Bests,
--
Edzer
On 01/27/2011 05:32 PM, Chris Mcowen wrote:
> Dear List,
>
> I apologise if this is not the correct place to ask this, however as ecologists' s i guess a high proportion of you would of had to deal with spatial autocorrelation.
>
> First,
>
> I have seen a method that used a GAM and included the coordinates of the grid cells ( sample locations, in my case this will be long&lat) as a smooth factor to account for spatial autocorrelation.
>
> I was wondering if anybody here had done this as i can not find a example, and am unsure how to structure the GAM.
>
> Second,
>
> Another spatial question -
>
>>> Spatial correlograms were constructed using Moran’s I coefficients at 10 distance classes.
>>
>
> Once again, has anyone done anything like this?
>
> Thanks,
>
> Chris
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
--
Edzer Pebesma
Institute for Geoinformatics (ifgi), University of Münster
Weseler Straße 253, 48151 Münster, Germany. Phone: +49 251
8333081, Fax: +49 251 8339763 http://ifgi.uni-muenster.de
http://www.52north.org/geostatistics e.pebesma at wwu.de
More information about the R-sig-ecology
mailing list