[R-sig-Geo] Error fitting separable space-time variogram model

rossiter_nl d.g.rossiter at utwente.nl
Tue Mar 11 16:07:17 CET 2014

Thank you for this help. Unfortunately the help page for fit.StVariogram 
gives no guidance on this, just "... arguments passed to optim". Looking at
?optim, we can see there are many control parameters, including bounds for
the box method "L-BFGS-B". But there is no mention of the parameters being
optimized in the various model forms (metric, separable etc.) and their
order, and no way to find out until the model is fit, then we can see the
attribute "pars" to see the order of parameters. The order in the call to
build the variogram model (which will be optimized) has the parameters in
the proper order (space range, then nugget; time range then nugget; sill)
once one realizes that the partial sills in the space and time models are
only proportional and will be adjusted according to the nugget. But that's
not obvious, it could be that the partial sills are adjusted and the nugget
follows along.

The examples in demo(stkrige) were very helpful; it might be good to include
a reference to that on the fit.StVariogram help page. Even better would be
to list the arguments to optim and their order for each model form. I did
not think to look at demo(package="gstat") to see if there were a demo on
space-time kriging, sorry for that.

Thanks for providing this nice functionality.

View this message in context: http://r-sig-geo.2731867.n2.nabble.com/Error-fitting-separable-space-time-variogram-model-tp7585910p7585925.html
Sent from the R-sig-geo mailing list archive at Nabble.com.

More information about the R-sig-Geo mailing list