Re: kriging with non-regular supports (polygons)

[Edzer J. Pebesma]

uleopold wrote:

On Thu, 21 Aug 2003, Edzer J. Pebesma wrote:

I tried to use the area option mentioned in appendix A.3. But it turned 
out to be extremely slow. Probably I made a mistake.

Below is my cmd file. It took gstat 2 hours to initialise the maps whereas 
it does take usually a few seconds. 'mask' and 'area map' are both raster 
maps on the same grid cell support. For the area map I chose integers 
classifying each polygon. Each polygon (area) has a different size. There 
are 31 in total.
 

I (think I) see what you would like. 'area' is a map that discretizes 
the different
areas. However, the command is 'area:...', not 'areas: ...'. Also you do 
specify
a mask map, so anything in the area map should be relative to (0,0), i.e. no
geographic coordinates, but coordinates relative to the cell centres in the
mask cells.

What I would do is this: do for every unit in 'LEI':

 - create an area map with only that unit non-missing valued
 - omit the 'mask' command, so the units' coordinates are absolute, not
    relative.
I agree that doing this without the do-loop would be nice :-)
--
Edzer