[R-sig-ME] LMER False Convergence (8)

[Ben Bolker]

Daniel Anderson <djauoregon at ...> writes:

>

 [snip]

> I tried installing the source version of lme4, as you suggested,
>  but still did not have success. I did get the
> following error message, however, so if the -3 
> version is significantly
> different than the -2 version I
> was able to install then it could still be an option. 
> 
> Warning: unable to access index for repository
> http://lme4.r-forge.r-project.org/repos/bin/windows/contrib/3.0
> 
>   There is a binary version available (and will be installed) but the
>   source version is later:
>          binary       source
> lme4 0.999999-2 0.99999911-3

  I've added a more recent binary; it may take up to 24 hours for the
r-forge repository to update.
If you're in a big hurry you can download the zip file from

https://r-forge.r-project.org/scm/viewvc.php/*checkout*/www/repos/bin/windows/
     contrib/3.0/
    lme4_0.99999911-3.zip?revision=1816&root=lme4

(URL broken to make Gmane happy, sorry)

and install it using repos=NULL

  The differences you're seeing below wouldn't worry me *very* much --
the results are certainly qualitatively similar -- but it would be
nice if the devel verion of lme4 happened to match HLM even more
closely (it's not guaranteed of course that HLM is right and lme4 is
wrong, but HLM has probably still been more thoroughly tested ...)
 

> 
> Thanks again for your help,
> Daniel
> 
> Parameter                            HLM                            lme4
> 
> Fixed Effects                                                      
> Intercept                         251.032824 (0.102705)   251.03099  (0.10443)
> Growth                               5.911175 (0.04071)     5.90961 (0.03776
> Variance Components                                                      
       
> Student intercept              41.38687                              41.51581
> Student slope                     0.50270                             0.45086
> Within student                   9.78905                              9.84352
> School intercept                 5.72659                              5.97047
> School slope                       0.85956                            0.72623
> 

  [snip]