Message-ID: <alpine.LRH.2.03.1307072134320.8732@reclus.nhh.no>
Date: 2013-07-07T19:45:03Z
From: Roger Bivand
Subject: SEM model : system is singular
In-Reply-To: <1372930688779-7584028.post@n2.nabble.com>
On Thu, 4 Jul 2013, erwin wrote:
> Hi,
>
> I have to do a spatial model for my master thesis. If I do a SEM or SAR
> model i get the message:
>
> Fehler in solve.default(asyvar, tol = tol.solve) :
> System ist f?r den Rechner singul?r: reziproke Konditionszahl =
> 3.61856e-20
>
> In other posts i have read that the reason could be multicollinearity. I
> have tested it with vif but all factors were below 2. I have also tried to
> add random numbers like suggested in another post but i got the same error
> message.
Instead of reading posts, which are often just guesses, do please read the
help page, which says that you may rescale the variables to avoid the
problem, whoich occurs when the spatial coefficient takes values that
differ greatly from sigma squared and the other coefficients.
>
> If i do an SAR with another method like "Matrix_J" i get the following error
> message:
>
> Fehler in as_dsTMatrix_listw(get("listw", envir = env)) :
> odd neighbours sum
>
Your weights appear not to be symmetric, which is a requirement for using
Cholesky decomposition. This should probably be trapped earlier; the "LU"
method should work for the same weights.
> So i don't get results at any type of model. I have a 2380*2380 weighting
> matrix could the size be a problem?
>
Not a problem on current hardware.
Hope this helps,
Roger
>
> Thank you very much in advance
>
> erwin
>
>
>
> --
> View this message in context: http://r-sig-geo.2731867.n2.nabble.com/SEM-model-system-is-singular-tp7584028.html
> Sent from the R-sig-geo mailing list archive at Nabble.com.
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
--
Roger Bivand
Department of Economics, NHH Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no