[GRASSLIST:5551] s.surf.rst cross-validation example

[José Gómez-Dans]

Hi!
I am interested in testing the cross-validation approach when using the
RST interpolation routine. I notice that in the 6.0beta there is a flag
to carry out cross-validation, but it is not entirely clear to me how
to use it. Is there some sort of tutorial on how to use it? 

In Neteler and Misatova's book, some software is mentioned, that uses a
jack-knife approach (it discards one point in the region, and checks
what the residual is). It also mentions using this procedure (and to
minimise the RMSE error over a region of interest) to estimate optimal
interpolation parameters (tension and smoothness). I presume that this
does not work "out of the box", but can be accomplished by a script
that tries interpolation over a parameter space and examining the RMSE
for each run, and deciding which one is smallest. If this is correct,
how does one do this? There is a cvdev output, but it's not documented
(at least, I can't seem to find it documented!).

Many thanks
José
-- 
Bristol Glaciology Centre
Geographical Sciences
University of Bristol, UK