[R-sig-ME] R-Forge lme4/lme4a installation on Windows 7

[Dennis Murphy]

lme4a is the experimental version of lme4 that resides on R-forge:
http://lme4.r-forge.r-project.org/. A significant chunk of its
functionality has not yet made its way into the production version
lme4 on CRAN. This is one reason why a number of mixed modelers prefer
to use lme4a; moreover, the draft of the lme4 book that Doug Bates
released last year, as well as  recent sets of his short course notes,
often uses code that exists in lme4a but not necessarily lme4.

AFAIK, the latest version of lme4a is one ahead of what I installed in
May on my Win7 box:
> packageVersion('lme4a')
[1] ‘0.999375.66’
I just checked on R-forge - the latest binary version of lme4a for
Windows is unavailable ['Page not found']. This is not unusual. The
version of lme4a I have installed works in conjunction with the
following development version of Matrix:
> packageVersion('Matrix')
[1] ‘0.9996875.0’
However, the production version of Matrix on CRAN is 0.999375-50
(which I also just checked). An error message occurs if one tries to
load the above version of lme4a on Windows with the production version
of Matrix. Consequently, on Windows at the present time, if you want a
working version of lme4a, you also need the development version of
Matrix from R-forge. If you want the CRAN version of Matrix (which
would make sense since numerous packages depend on it), then you're
out of luck with lme4a. [I'm happy to be corrected if this information
is out of date, since I encountered this problem in May.]

 I've run into at least one situation where the development version of
Matrix clashes with another package: doBy. This is unfortunate since
Soren Hojsgaard recently added new functionality to the package that
pertains to mixed model analysis.

The package developers use Debian Linux, so the safest way to use the
latest and greatest version of lme4a is to work on a Linux platform.
Windows [and Mac?] users need to be more patient and flexible.

It just occurred to me that a possible remedy to this problem may be
found in the recently released devtools package by Hadley Wickham. He
wrote it so that users could load and test the development version of
ggplot2 (a significant rewrite of the package) without clashing with
the existing production version. It contains a function dev_mode()
that one calls to toggle between the production and development
versions of a package, but you do need to install (and thus flag)
development versions while devtools is loaded and in dev_mode. I
didn't see anything specific about installing from R-forge, but it
shouldn't be all that difficult to implement since the package was
written for general use. Just a suggestion to pass on to the
development team...

Dennis


On Thu, Jul 21, 2011 at 7:32 AM, David Winsemius <dwinsemius at comcast.net> wrote:
>
> On Jul 21, 2011, at 4:08 AM, Dieter Menne wrote:
>
>> Simon Delisle-Boulianne <simondelisleboulianne at ...> writes:
>>>
>>> I don't know if I'm the only one, but I'm unable to install lme4 or lme4a
>>> (or Matrix for that matter) from R-Forge since a couple of weeks (
>>> http://r-forge.r-project.org/R/?group_id=60). I'm running R version
>>> 2.13.1
>>> on windows 7 (64 bits). I get the following error message:
>>>
>>> *> install.packages("lme4", repos="http://R-Forge.R-project.org")*
>>> Warning : unable to access index for repository
>>
>> The Windows build has not been created since some time.
>>
>> Ben Bolker had an alternative at:
>>
>> install.packages("lme4a",repos="http://www.math.mcmaster.ca/bolker/R")
>>
>> but as far as I see it is a 2010 version.
>>
>
> What was puzzling to me was why r-forge was being used at all? lme4 is a
> regular CRAN package and has Windows binaries there ... as is typical.
>
> This is what one sees at r-forge:
> https://r-forge.r-project.org/forum/forum.php?forum_id=651
>
> And Matrix is part of the regular R distribution, so why would its absence
> on r-forge be an issue?
>
> --
> David Winsemius, MD
> West Hartford, CT
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>