David, I will Cc: this to gstat-info, which is the place for questions of this kind.
While fitting you probably ignored the warning:
Warning: singular model in variogram fit
which means: I cannot fit a sensible model here. I will change gstat in the near future such that it does not return a variogram model after this message; you'd better ignore it.
If you insist on fitting a double exponential model, you may want to fix one or more of its coefficients, starting with range values.
The two initial models you start with are very similar, so for any non-linear fit routine it is hard to decide where to go from there. -- Edzer
David Hetherington wrote:
Dear Edzer
We are trying to fit nested exponential models in gstat (R add on version). The problem we are facing is that the least square fitting algorythm (method #6) does not seem to be converging to sensible parameter estimates, plotting the model against the variogram it is easy to see the fit is really not good. For example
davidnp dist gamma dir.hor dir.ver id
1 2337544 4.158366 224.5603 0 0 1
2 6912412 9.669306 481.9654 0 0 1
3 11031598 15.784569 723.1225 0 0 1
4 14854496 21.980985 875.0107 0 0 1
5 18455560 28.233768 873.8049 0 0 1
6 21232724 34.474862 848.9212 0 0 1
7 23888652 40.671650 819.0625 0 0 1
8 26675362 46.917612 809.4374 0 0 1
9 28458876 53.172072 813.5839 0 0 1
10 30503760 59.426193 832.5799 0 0 1
11 31633408 65.677430 851.3040 0 0 1
12 33125978 71.928246 864.0765 0 0 1
13 33684580 78.177352 862.5658 0 0 1
14 34383872 84.407302 848.3167 0 0 1
15 34783728 90.654143 825.8910 0 0 1
model.1 <- fit.variogram(david,vgm(600,"Exp",30,add.to=vgm(5,"Exp",20)))
plot(david, model.1)
model.1
model psill range 1 Exp 328.5582 4156.3973 2 Exp 604.0956 32.0286
as you can see, the first component of the model is hardly contributing to the semi-variance because the range is orders of magnitude too large, but the second component also fails to fit the points of the semi-variogram. How can we adapt the gstat commands such that the least squares algorythm converges on a meaningful answer?
Thanks in advance David Hetherington and David Pleydell
David Hetherington Telford Institute of Environmental Systems, School of Environment and Life Sciences, Peel Building, University of Salford, Salford, M5 4WT
Telephone: 0161 2953261 Facsimile: 0161 2955015 http://www.ties.salford.ac.uk/pg/dh