variogram.fit(vobj, param, fun=spher.vgram, lower=rep(0, n.param),
upper=Inf)
"variogram" representing
an empirical variogram. Usually,
the result of the
variogram function.
"nugget",
"sill", and
"range"
or a subset of these.
If missing, the function will try to determine the parameter names
and initial values based on the arguments to the function specified in
fun.
nlminb for more information.
nlminb for more information.
"vgram.fit" with components:
fun argument as a character string.
vobj.
If fun is one of
exp.vgram,
gauss.vgram,
linear.vgram,
power.vgram
or
spher.vgram and
param is not supplied the
function sets special initial starting values for
param.
Otherwise, if
param is not supplied it is set to a vector of
ones.
The weighted least squares objective function used in the fitting process
(Cressie, 1993, p. 97) is:
objective.fun <- function(y,yh,n) sum(n*(y/yh-1)^2)
The
nlminb function is used for the optimization.
Cressie, Noel. (1993). Statistics For Spatial Data, Revised Edition. Wiley, New York.
vg.iron <- variogram(residuals ~ loc(easting, northing), data=iron.ore)
vfit.iron <- variogram.fit(vg.iron, param=c(range=8.7, sill=3.5,
nugget=4.8), fun=spher.vgram)
plot(vg.iron)
plot(vfit.iron, add=T)