[R] Multiple CHOLMOD errors when attempting poisson glmm

[Douglas Bates]

On Thu, Dec 24, 2009 at 1:03 PM, postava-davig.m
<postava-davig.m at husky.neu.edu> wrote:
>
> Hello,

> I have been attempting to run a poisson glmm using lme4 for some time now
> and have had a lot of trouble.  I would say 9 times out of 10 I receive the
> following warning:

> CHOLMOD warning:  %h
> Error in mer_finalize(ans) :
>  Cholmod error `not positive definite' at
> file:../Cholesky/t_cholmod_rowfac.c, line 432

That is an (admittedly obscure) indication that the Cholesky
factorization of a matrix derived from the random-effects model matrix
cannot be performed.

> My data are counts of microbe colony forming units (CFUs) collected from
> termite cuticles and the surrounding environment over a 3 year period.  I am
> attempting to analyze the effect of several factors on these counts (termite
> nest volume, temperature, humidity, light, incubation temperature, habitat,
> year, sample location, etc.) to determine which account for the variance in
> microbial communities.  These data are observations, so there are many
> missing values....which may be part of the problem.  I've tried many
> different combinations of variables, and also have tried reducing my data
> set to remove as many NA's and confounding variables as possible, but I
> still can't get any models to work consistently.  One most recent attempt
> had the following output:

 model1=lmer(totalcfus~habitat*temp*moisture*light+location+(1|habitat/colony/location),family=poisson,control=list(msVerbose=1))
>  0:     553377.59:  1.00573 0.620530 0.169516  26.3904 -13.1266 -33.2286
> -21.1955 -21.1064 -0.590761 -0.217403 -0.0342272 -0.960593 -0.0962517
> 0.441626  1.20575 0.718621 0.680580 0.171006 0.403729 0.278822 0.275395
> 0.00707767 0.0225599 0.0854869 0.0533373 0.0243451 0.00114120 0.000403226
> -0.00566960 -0.0143715 -0.00931896 -0.00879323 -0.000753236 -0.00335745
> -0.00178054 -0.000788027 -0.000288944 -0.000909455 -0.000839295 -0.000309293
> -1.35885e-05 9.76120e-06 3.57035e-05 2.78985e-05 1.01880e-05
> CHOLMOD warning:  %h
> Error in mer_finalize(ans) :
>  Cholmod error `not positive definite' at
> file:../Cholesky/t_cholmod_rowfac.c, line 432

Thank you for including the output from verbose = TRUE.  It would also
help if you included the output from sessionInfo() so we can see which
version of R you are using and which version of the lme4 package you
are using.

How many observations are used in this fit?  As you can see, the
number of parameters being fit is very large and encountering
singularities is not unexpected.

May I suggest that we move this discussion to the
R-SIG-Mixed-Models at R-project.org mailing list, which I have cc:d on
this reply?  That list is specifically intended for discussions of
this type.
> I have to admit that I'm at a loss, and have been unable to determine any
> pattern to when this error message comes up.  I'm hoping that someone can
> help me eek out what the issue is with my data so that I can eventually work
> out a usable model.
>
> Thanks so much, and happy holidays.
> --
> View this message in context: http://n4.nabble.com/Multiple-CHOLMOD-errors-when-attempting-poisson-glmm-tp978573p978573.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> 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.
>