[R] SEM error

[John Fox]

Dear Jarret, Uwe, and Dan,

Sorry -- I missed the initial question. What's a bit odd here is that the singularity occurs only in the computation of the modification indices. It might help to look at the conditioning of the covariance matrix of the parameter estimates (i.e., the eigenvalues or singular values), which must have been invertible.

Regards,
 John

On Mon, 22 Feb 2010 09:00:08 -0800
 Jarrett Byrnes <byrnes at msi.ucsb.edu> wrote:
> I have often found this to happen if the scale of one variable is  orders of magnitude different than the scale of other variables.  Have  you tried inspecting the covariance matrix and log transforming any  such variables?
> 
> On Feb 22, 2010, at 8:14 AM, Uwe Ligges wrote:
> 
> >
> >
> > On 20.02.2010 08:51, Dan Edgcumbe wrote:
> >> I'm trying to do some confirmatory factor analysis on some data. My  
> >> SEM
> >> model solves in 22 iterations, but when I try to look at the  
> >> modification
> >> indices, using mod.indices, I get the following error message:
> >>
> >> Error in solve.default(hessian) :
> >>   system is computationally singular: reciprocal condition number =
> >> 4.40283e-18
> >>
> >> What does this mean?
> >
> > That the method you apply tries to invert some object called  
> > "hessian" (maybe a hessian? ;-)) but fails since a singular matrix  
> > cannot be inverted. Perhaps (as I often found for people doing sem  
> > analyses) you have less observations than parameters to estimate or  
> > only certain combinations for some factors?
> >
> > Uwe Ligges
> >
> >
> >
> >
> >
> >>
> >> Many thanks,
> >>
> >> Dan
> >>
> >> 	[[alternative HTML version deleted]]
> >>
> >> ______________________________________________
> >> R-help at r-project.org mailing list
> >> https://stat.ethz.ch/mailman/listinfo/r-help
> >> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> >> and provide commented, minimal, self-contained, reproducible code.
> >
> > ______________________________________________
> > R-help at r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> > and provide commented, minimal, self-contained, reproducible code.
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

------------------------------------------------
John Fox
Sen. William McMaster Prof. of Social Statistics
Department of Sociology
McMaster University
Hamilton, Ontario, Canada
http://socserv.mcmaster.ca/jfox/