
Hi,
Thanks for the answer on my previous question. However, it leads me to new
problems.
I am struggling to bound the relative tausq parameter to not be too high
in my
estimations. I do not want it to be above a certain limit that would
violate the
model; it yields surfaces that are too ruff. First I want estimates from
variofit,
but this does not seem to work. Shouldn't this give a tausq^2/sigmasq^2
in (0,0.05)?
library("geoR")
data(s100)
vario100 < variog(s100, max.dist=1)
ini.vals < c(1,0.5)
limits < pars.limits(phi=c(0,2),nugget.rel=c(0,0.05))
ols < variofit(vario100, ini=ini.vals, fix.nug=FALSE,
limits=limits,wei="equal")
ols$nugget/ols$cov.pars[1]
Thanks in advance,
Anders Malmberg
