Question about correcting for spatial autocorrelation

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Question about correcting for spatial autocorrelation

R-sig-geo mailing list
Hi 
I've posted this to the mixed models forum as well but thought people here might know:
I want to use the the correlation setting with corSpher in nlme to account for potential spatial autocorrelation in my data. My data include observations from across the globe with locations in latitude and longitude (decimal degrees). From the R help for corSpher, the syntax for their wheat example would be something like:
fm1Wheat2 <- gls(yield ~ variety - 1, corr =corSpher(c(28, 0.2), form = ~ latitude + longitude, nugget = TRUE))

My question is whether the latitude and longitude provided should be projected into a spatial projection that preserves distances or areas or whether providing decimal degrees is appropriate?
Many thanks!
        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Question about correcting for spatial autocorrelation

Roger Bivand
Administrator
On Fri, 22 Jun 2018, Julie Lee-Yaw via R-sig-Geo wrote:

> Hi 
> I've posted this to the mixed models forum as well but thought people
> here might know:
> I want to use the the correlation setting with corSpher in nlme to
> account for potential spatial autocorrelation in my data. My data
> include observations from across the globe with locations in latitude
> and longitude (decimal degrees). From the R help for corSpher, the
> syntax for their wheat example would be something like:

Nothing in the documentation nor the underlying book suggests that the
coordinates are anything other than planar, so geographical coordinates in
decimal degrees should not be used. Unfortunately, I do not have access to
the underlying article in crop science, but there is no reason to think
that the agricultural experiment extended from France to Turkey and France
to south of Ghana. The documentation in nlme::corSpher() refers to
stats::dist, and I can't see any way to insert fields::rdist.earth() in
its place. So you can safely accept that the coefficients must be
projected (i.e. planar). This is all accessible in the documentation and
Pinheiro & Bates (2000). It is pretty certain that they did not envisage
global scope in geographical coordinates, so maybe you should be using
fields or gstat which are more likely to be able to handle Great Circle
distances.

Hope this clarifies,

Roger

> fm1Wheat2 <- gls(yield ~ variety - 1, corr =corSpher(c(28, 0.2), form =
> + ~ latitude + longitude, nugget = TRUE))
>
> My question is whether the latitude and longitude provided should be
> projected into a spatial projection that preserves distances or areas or
> whether providing decimal degrees is appropriate?

> Many thanks!

Please post plain text.

> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> [hidden email]
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

--
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: [hidden email]
http://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en
_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Roger Bivand
Department of Economics
Norwegian School of Economics
Helleveien 30
N-5045 Bergen, Norway