R help - Sep 2003 - glmmPQL probelm

Nothing to do with glmmPQL (as traceback() would have shown you).

I think you are looking for ?try.

On Wed, 3 Sep 2003, Andrej Kveder wrote:
> Dear listers,
> 
> First let me appologize if the same mail arrives multiple times. Recently I
> had some probelms sending my e-mails to the list.
> 
> I encountered a problem when running glmmPQL procuedure doing multilevel
> modeling with a dichotomous outcome.
> Those are the two error messages I usually get:
> 
> Error in logLik.reStruct(object, conLin) :
>         NA/NaN/Inf in foreign function call (arg 3)
> 
> Error in solve.default(pdMatrix(a, fact = TRUE)) :
>         Lapack routine dgesv: system is exactly singular
> 
> The trick is that the model is a part of a simualtion run, which uses the
> same starting variance covariance matrix as a source for a mutlivariate
> normal simulated 2 level dataset. So the variations in the data set are
just
> a part of the stochastic process. In the majority of the cases the model
> runs fine, while in some cases I get either of the two error messages. The
> model obviously cannot be identified. Does anybody have an idea why?
> The second question would relate to the handling of the problem. I
don't
> even mind that much to have some models unidentified (of cours I would
> rather not have them). I would like to detect the possible problem
> beforehand in order to instruct the simulation run not to specifiy it. What
> is most annoying is that when an error occurs, the program stops the
> simulation run and not just skips the model results. Does anybody have an
> idea what to do about it? Any suggestions would be most helpfull. I can
also
> send the R program file if that would help.
> 
> Thanks
> 
> Andrej
> 
> _________
> Andrej Kveder, M.A.
> researcher
> Institute of Medical Sciences SRS SASA; Novi trg 2, SI-1000 Ljubljana,
> Slovenia
> phone: +386 1 47 06 440   fax: +386 1 42 61 493
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> 
> 
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Dear listers,

First let me appologize if the same mail arrives multiple times. Recently I
had some probelms sending my e-mails to the list.

I encountered a problem when running glmmPQL procuedure doing multilevel
modeling with a dichotomous outcome.
Those are the two error messages I usually get:

Error in logLik.reStruct(object, conLin) :
        NA/NaN/Inf in foreign function call (arg 3)

Error in solve.default(pdMatrix(a, fact = TRUE)) :
        Lapack routine dgesv: system is exactly singular

The trick is that the model is a part of a simualtion run, which uses the
same starting variance covariance matrix as a source for a mutlivariate
normal simulated 2 level dataset. So the variations in the data set are just
a part of the stochastic process. In the majority of the cases the model
runs fine, while in some cases I get either of the two error messages. The
model obviously cannot be identified. Does anybody have an idea why?
The second question would relate to the handling of the problem. I don't
even mind that much to have some models unidentified (of cours I would
rather not have them). I would like to detect the possible problem
beforehand in order to instruct the simulation run not to specifiy it. What
is most annoying is that when an error occurs, the program stops the
simulation run and not just skips the model results. Does anybody have an
idea what to do about it? Any suggestions would be most helpfull. I can also
send the R program file if that would help.

Thanks

Andrej

_________
Andrej Kveder, M.A.
researcher
Institute of Medical Sciences SRS SASA; Novi trg 2, SI-1000 Ljubljana,
Slovenia
phone: +386 1 47 06 440   fax: +386 1 42 61 493

	[[alternative HTML version deleted]]