[R-sig-ME] "invalid class “mer” object" error with lme4 0.999375.42

[Ben Bolker]

On 13-11-23 07:50 PM, Rachel Smith wrote:
> Dear lme4-folks,
> 
> I wonder if you can help me with a query.
> 
> In our lab we are (not very advanced) users of lme4, who have until
> recently relied heavily on languageR. We are trying to adjust to lme4
> 3.0.1 and life without languageR/pvals.fnc.

  Sorry about that :-(

> As part of this process, to try to ensure continuity for our
> students, I recently asked our technical support to ensure that our
> lab machines are running the following: R 2.15.2 with lme4 version
> 0.999375.42, and languageR R 3.0.1 with lme4 version 1.0.5
> 
> I have tested the following simple bit of code under both of the
> above. junk = lmer(RTlexdec ~ WordCategory + (1|Word), data=english) 
> This works fine under R 3.0.1 with lme4 1.0.5.
> 
> But, under R 2.15.2 with lme4 0.999375.42, I get the following error
> message:
> 
> "Error in validObject(.Object) : invalid class “mer” object: Slot L
> must be a monotonic LL' factorization of size dims['q']"
> 
> 
> 
> Do you have any idea where this problem could be originating? I'm
> wondering whether the versions of e.g. MASS or matrix that we have
> loaded, could be an issue here - might we have newer versions of
> these that aren't compatible with lme4 0.999375.42, or something like
> that?
> 

   Try re-installing lme4.0 *after* you have installed the new versions
of Matrix; if you switch Matrix versions "under" lme4.0 , it won't
detect that the binary interfaces have changed and that it should
therefore be re-installed ... (i.e. it's not technically that the newer
versions of Matrix are incompatible, but that lme4.0 needs to be
recompiled/reinstalled to work properly with the new version)

  Ben Bolker