From: Edzer J Pebesma <e.pebesma at geo.uu.nl>
To: Al? Santacruz <amsantac at hotmail.com>
CC: r-sig-geo at stat.math.ethz.ch
Subject: Re: [R-sig-Geo] ossfim with Universal Kriging?
Date: Sat, 08 Jul 2006 21:48:21 +0200
Ali, if you replace e+n with x+y, does that work?
I wasn't aware that ossfim was this restrictive; there's
clearly some room for generalisation here.
--
Edzer
Al? Santacruz wrote:
Dear list members:
I am trying to make a cost/benefit analysis in a sampling network for
organic matter soil content data. The best model fitted was exponential
(universal kriging, first order trend).
So I need to know if it is possible to use ossfim function with universal
kriging.
I tried to modify the ossfim function but I get an error:
ossfim.UK<-function (spacings = 1:5, block.sizes = 1:5, model, nmax = 25,
+ debug = 0)
+ {
+ n = floor(sqrt(nmax)) + 1
+ x = 0:(n - 1) + 0.5
+ x = sort(c(-x, x))
+ ret = matrix(NA, length(spacings) * length(block.sizes),
+ 3)
+ r = 1
+ for (sp in spacings) {
+ for (bl in block.sizes) {
+ data.grid = data.frame(expand.grid(x * sp, x * sp),
+ z = rep(1, length(x)^2))
+ names(data.grid) = c("x", "y", "z")
+ gridded(data.grid) = c("x", "y")
+ x0 = SpatialPoints(matrix(0, 1, 2))
## Next I modify the formula in krige using the coordinates names of my
data
## to use universal kriging:
+ kr = krige(z ~ e+n, data.grid, x0, block = c(bl, bl),
+ model = model, nmax = nmax, set = list(debug = debug))
+ ret[r, ] = c(sp, bl, sqrt(kr[["var1.var"]][1]))
+ r = r + 1
+ }
+ }
+ ret = data.frame(ret)
+ names(ret) = c("spacing", "block.size", "kriging.se")
+ ret
+ }
err.grid <- ossfim.UK(seq(10,60,10),block.sizes=0,m.f.exp.t1)
Error in model.frame(formula, rownames, variables, varnames, extras,
extranames, :
variable lengths differ (found for 'e')
Is it possible what I want to do? Does it have sense?
I appreciate any help or comment.
Sincerely,
Al? M. Santacruz
M.Sc. Geomatics, Student
National University of Colombia
Bogot? D.C.