[R-sig-Geo] corSpatial and zero distance

[Edzer Pebesma]

lme fits using (restricted) maximum likelihood, and that requires a
covariance/correlation matrix that is non-singular, to be inverted each
iteration.

Under spatial correlation models, even the pure nugget model,
observations with identical location are perfectly correlated (as in:
the correlation of X with itself), and so result in duplicate rows/cols,
making the covariance/correlation matrix singular.

On 10/31/2010 11:21 AM, valerio.bartolino at uniroma1.it wrote:
> Dear list,
> it is probably a very simple question with an obvious answer, that unfortunately I cannot find by myself.
> 
> Why I cannot fit a spatial correlation structure model if I've some observations in the same location? Shouldn't the nugget account exactly for this small-scale variability and measurement errors?
> 
> I'm using one of the correlation models in the generic function 'lme' (package:nlme)
> 
> Thank you
> 
> Valerio
> 

-- 
Edzer Pebesma
Institute for Geoinformatics (ifgi), University of Münster
Weseler Straße 253, 48151 Münster, Germany. Phone: +49 251
8333081, Fax: +49 251 8339763  http://ifgi.uni-muenster.de
http://www.52north.org/geostatistics      e.pebesma at wwu.de