[R-sig-Geo] krige.cv, again :)

pebesma at geo.uu.nl pebesma at geo.uu.nl
Sat Oct 7 12:04:07 CEST 2006

> hi, me again :)
> I've read that in the help of krige.cv :
> Note:
>      Leave-one-out cross validation seems to be much faster in plain
>      (stand-alone) gstat, apparently quite a bit of the effort is spent
>      moving  data around from R to gstat.
> Since I have 6 ensemble of 2300 stations, I was wondering : How could I
> launch gstat in stand-alone?  Do I need unix or it's possible to run it on
> Windows?
Either of the two; windows binary is at http://www.gstat.org/
> At this moment, I'm running R 2.3.1 on a windows machine with 512MB of RAM
> and a P4 2.6 GHz, and it's impossible to do a leave-one-out approach that
> takes less than 2 days to run.
Do you use kriging in a local neighbourhood? If not, moving out of R may
not help much, as 2300 times solving a new system of size 2299 will be the
bottle neck.

More information about the R-sig-Geo mailing list