Rebecca wrote:

I might be really daft but is there an automatic way of getting the
goodness-of-fit statistics for fitted variograms?

The answer is, at least I believe, and correct me if I am wrong, no.
This option was not included in the Gstat program on purpose because
goodness-of-fit statistics can give the misleading impression of quality
when they are high, but the choice of the model is in reality rather
subjective.
A method that you can apply to get the goodness-of-fit statistics - and
that is normally applied at this institute - is to program the model
that you fitted in a spreadsheet, read off the values of the
experimental variogram in the Gstat variogram screen, note them in your
spreadsheet as well and let the spreadsheet calculate the
goodness-of-fit in the form of a r square (this is rather easy in Excel
with the use of line fitting in a graph and selection of 'Display R
squared', and in Quattro Pro using the menu Extra with Statistical
functions).
Instead of programming the model in the spreadsheet, you can also read
off the values for the variogram model, put them next to the
experimental variogram values, and let the spreadsheet calculate the r
square, but this is of course even less precise - or you have to use a
ruler on your computer screen - than the above mentioned method. At
least it gives you an impression of the goodness-of-fit and you can
compare roughly the different fits for different models.

With greetings,

Ivo Thonon
--
Laboratorio de Edafolog�a
Facultade de Ciencias
Universidade da Coru�a
Campus da Zapateira
15.071  A Coru�a
Galicia, Espa�a


Reply via email to