[R-sig-ME] spatial correlation ANOVA-GLS

[Ing. Agr. Agustín Alesso]

Hello,

I need some advice about how to include a spatial correlation structure
to an ANOVA-GLS one-way model. 

After fitting a simple model asuming uncorrelated errors and plotting a
variogram with its normalized residuals variogram the spatial
autocorrelation was evident (nuget 0,5 and range 600 m aprox).
So I tried to fit few models with predefined corStructures (lin, exp,
sph) without initial values for range and nugget and then looking at
their normalized residuals variograms. Although including correlation
results in smaller AICs, the variograms still showing strong spatial
correlation of normalized residuals.  

1) Could it be because of a poor estimation of corStructure? Indeed With
gstat package, the best model fitted has a nugget efect with to
spherical structures (2 ranges).

2) Is there a way to bluid a spatial correlation structure including two
ranges?

I'd appreciate some tip. Thanks.


-- 
Ing. Agr. Agustín Alesso
Dpto. Ciencias del Ambiente
Fac. Cs. Agrarias - UNL
Kreder 2805 - S3080HOF - Esperanza
Tel: 03496-428575 int 337/256