[R-sig-ME] lmer() on linux

[Douglas Bates]

On Thu, May 6, 2010 at 10:34 AM, Julia Liu <liujulia7 at gmail.com> wrote:
> Dear Prof. Bates,
>
> I have been using lmer() for sometime in windows version of R. Recently, I
> decided to install R on a linux machine in order get bigger memory limit.
> Everything went fine, including testing most of my existing R codes.
> However, I have ran into one problem with function lmer() on Linux R. For
> some of my data and codes, the lmer() gave the following warning:
> "Warning message:
> In mer_finalize(ans) : iteration limit reached without convergence (9)"
>
> And the exact same code and data ran fine in Windows R (without any
> warnings).  What's is the reason? Does Linux version have a lower iteration
> limit than the windows version?

Even on the same hardware there are subtle differences in the
numerical properties of some of the optimization and linear algebra
methods between Windows and Linux.  This warning comes from the
optimization method.

Could you provide a reproducible example, including the data?  If the
data are proprietary you could remove any identifying characteristics
and anonymize the identifiers.  I would just need to have the
numerical values of the covariates and response plus the formula
defining the model.

It is generally better to send questions like this to the
R-SIG-Mixed-Models at R-project.org mailing list, which I have taken the
liberty of cc:ing on this reply.  I am deep in the throes of writing
and debugging a new version of the code and don't always have the
opportunity to respond quickly to inquiries.