[R-sig-Geo] Thanks to all: Distances in geoR package

[Roger Bivand]

On Fri, 17 Jun 2005, Paulo Justiniano Ribeiro Jr wrote:

> A few words more on this.
> 
> My feeling is that the best way to handle this is with coordinate
> trasformation (for instance for UTM) and then backtransform
> at the  end if required.
> 

And a couple more ...

Which way to go depends on the relative scale of the phenomena of 
interest. For up to hundreds of km, the distortions introduced by 
projecting to UTM or other suitable projections are small. But at 
continental scales, working with Great Circle distances for building 
distance matrices is a much better choice, not depending on the origin of 
the projection. 

I've had C code in spdep for some time to do GC (an if() on a flag passed
through .Call to choose between the R built-in pythag() or the GC
distance). I've put a copy in the spgwr package, but multiple copies means
maintenance problems. The rdist.earth() function in fields is at the R
level - the nice comment on the help page does hold: "Surprisingly this
[is] all done efficiently in S".

I'd like to expose the GC function in the sp package, but have not yet
looked at how contributed packages can share functions (and header files)
at the C level - the header file is the real issue, I think. Would there
be interest in doing this? The GC function as it is does not have an R
wrapper, because in spdep, most of the distances are thrown away
immediately as being beyond threshold.

Feedback, please?

Roger

> There were a few messages along these lines and I will take the
> opportunity to mention that there is a package under development (aRT)
> which intends to provide this and other functionalities typical from GIS
> systems via Terralib
> The package page is: http://www.est.ufpr.br/aRT
> 
> Still changing rapidly but already some functionalities are already
> working.
> There is a package vignette and a few exemples in the tests directory.
> This package adheres the the "sp" classes andin newer versions also allows
> "proj4".
> 
> Best
> P.J.
> 
> 
> 
> On Fri, 17 Jun 2005, giovanna jona lasinio wrote:
> 
> > Dear all, thanks for the many suggestions, I'll try and let you know
> > what worked best. Package fields works well in computing variogram's
> > cloud with lon-lat coordinates (returns by default distances in miles)
> > and usually I use a loess-built curve to get a nice-looking version of
> > the variogram. However I wasn't able to use the "vgram" function's
> > output in order to obtain a "variofit" estimated model. What would be
> > extremely usefull, in my opinion, is to allow "variog" to accept a
> > distance matrix as input, in this way no coordinate transformation would
> > be needed. I'm not yet good enough at programming in R to find a good
> > solution myself :-) but if I find a solution I'll let you know....
> >
> > Best
> > Giovanna Jona Lasinio
> >
> > _______________________________________________
> > R-sig-Geo mailing list
> > R-sig-Geo at stat.math.ethz.ch
> > https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> >
> >
> 
> Paulo Justiniano Ribeiro Jr
> LEG (Laboratório de Estatística e Geoinformação)
> Departamento de Estatística
> Universidade Federal do Paraná
> Caixa Postal 19.081
> CEP 81.531-990
> Curitiba, PR  -  Brasil
> Tel: (+55) 41 3361 3573
> Fax: (+55) 41 3361 3141
> e-mail: paulojus at est.ufpr.br
> http://www.est.ufpr.br/~paulojus
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no