Skip to content

Note on variogram fitting with polynomial degree

3 messages · Caspar Hallmann, Edzer Pebesma

#
Dear all,

I read in ASDAR: "Residual variograms are calculated by default when a
more complex model for the trend is used..."

However, I noticed that when using fit.variogram in gstat with an
additional option: degree=2, but keeping the formula as an intercept
only, does not account for the higher degree of complexity in the
model. I suspect this is not the intended behavior. To honour a
dependence structure (polyn degree) in my observations I need to state
the dependence explicitly in the formula.

Best regards,

Caspar

data(meuse)
coordinates(meuse) <- c("x", "y")
v<-variogram(log(zinc)~1,meuse)
print(v)
vs<-variogram(log(zinc)~1,meuse,degree=2)
print(vs)
vt<-variogram(log(zinc)~x*y+I(x^2)+I(y^2),meuse)
print(vt)

PS:
x86_64-w64-mingw32/x64 (64-bit)
R version 3.0.1 (2013-05-16)
gstat_1.0-16
#
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

degree is not listed as a parameter in ?variogram, and is ignored,
regrettably silently, but this is how R often works when there's a ...
in the argument list.

The function that accepts degree is ?gstat, and I tried:

variogram(gstat(NULL, "lzinc", log(zinc)~1, meuse, degree=2))

and that gives a nice error message that degree is not supported by
variogram().

I understand this is somewhat confusing. Do you think we should modify
the book or the documentation of some function?
On 09/17/2013 11:36 AM, Caspar Hallmann wrote:
- -- 
Edzer Pebesma
Institute for Geoinformatics (ifgi), University of M?nster
Heisenbergstra?e 2, 48149 M?nster, Germany. Phone: +49 251
83 33081, Fax: +49 251 8339763 http://ifgi.uni-muenster.de
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJSOC6KAAoJEM1OCHCtOnfx6dkH/iIluvBI83ghDMoNIb+/rTDz
LOtVwu8oZp90vGIAVltBbqPVIu4SZH4jj6jif0HrEJU3wshE4k51EkGwq77jvP13
+pnCrwApeBHpJ2iUgYgqYCJAkwj+F25HxX1IxjcmQVIe8LC6O3fl/Rr8WAdnqbAL
M3J8nzEf23yWba683BCBxQwDBHbaTSHYlrKjAbEyQWsnVY40lBV0Vl7SdooseyWK
jPdEFPhp3kQIpPKhbmoAo9PwocwfUympZZ0KbM3xosZrI2cfI1dXXzrOwJtsCBrW
7/pK9fOxDYfGZICEnSr3xkf3FQCaoFQZY5uVEIrOo67/GkNxt5ewBkIdsQCi+vc=
=m5bR
-----END PGP SIGNATURE-----
#
Hi Edzer,
Thanks for the clarification. Well, ideally, I think variogram()
should be made possible to account for "degree". Alternatively a
simple note in the help file for variogram along the lines "NOTE:
degree is not used in variogram(), Use formula to specify a trend
surface instead", might suffice.
Thanks again
Caspar


On Tue, Sep 17, 2013 at 12:27 PM, Edzer Pebesma
<edzer.pebesma at uni-muenster.de> wrote: