[R-sig-Geo] 2 problems with errorsarlm(method="SparseM")

[Martin Reismann]

Roger kindly provided solutions for both problems.
I write a short summary for the interested public:

> (1) When running a large SAR (weight matrix 12544 * 12544) with errorsarlm
> and SparseM-method I ran into this error message:
>
> > Error in .local(x, ...) : Increase nnzlmax
> > Error in det(chol((I - lambda * csrw), tmpmax = tmpmax)) :
> >         unable to find the argument 'x' in selecting a method for function 'det'

Quoting Roger Bivand:
> After making changes, the ML functions have a new argument allowing the
> memory allocation arguments for sparse Cholesky can be passed through and
> the defaults have been made larger.
> ...
> I'm attaching a draft source package for you to try out - see the help
> page for details on passing the arguments through cholAlloc=; in the
> defaults I've used multiples of sum(card(listw$neighbours)), the number of
> non-zero elements in the matrix.

The draft source package is spdep_0.3-20.tar.gz and most likely available
from Roger.


> (2) Being able to use mat2listw instead of nb2listw would be preferable
> because mat2listw allows matrix elements different from 0 and 1.

the glist= argument to nb2listw() allows to assign a weight to every
2-tuple of neighbours. Thus it can be used instead of mat2listw() for
preparing listw-objects for errorsarlm.

--
Mag. Martin Reismann
c/o
Institut fuer Wirtschaftsgeographie und Geoinformatik
Wirtschaftsuniversitaet Wien
Nordbergstrasse 15, 4. Stock, Kern A
A-1090 Wien

EMail: Martin.Reismann at wu-wien.ac.at
Tel +43-(0)-1-31336-5720
Fax +43-(0)-1-31336-703