[R] Bugs through R in Mac

Uwe Ligges ligges at statistik.tu-dortmund.de
Tue Jan 22 09:56:24 CET 2008

Bill Northcott wrote:
> On 21/01/2008, at 10:00 PM, Uwe Ligges wrote:
>> Fredrik Lundgren wrote:
>>> Hello,
>>> I recently changed from Win XP to Mac OS X (10.5.1).
>>> Is there a way to run  Bugs (in any version) in R (R version 2.6.0
>>> (2007-10-03)) on this platform?
>> Which way to "run Bugs in R" did you use in Windows?
>> On MacOS X, a solution that should work but is completely inelegant is
>> using the package R2WinBUGS, and running WinBUGS within DarWINE. There
>> are some rumors that OpenBUGS development will be forced within the next
>> couple of months, but I am not that confident that we will have a
>> natively working BRugs version on Mac OS X or Linux this year ...
> I just posted a message about this on r-sig-mac.
> I spent a little time on this late last year.
> WinBUGS runs happily under DarWINE or the commercial package 
> CrossoverMac.  There is nothing inelegant about it.

It is inelegant in at least two points:
1. R2WinBUGS is a mess per se, we all want to use something that runs 
natively within R, such as BRugs or jags.
2. running the WinBUGS instance in xyz-WINE things does not make things 
faster and more elegant from my point of view.

> R2WinBUGS is another can of worms.  The code to support running WinBUGS 
> in WINE on Linux or Mac is thoroughly broken.  The problem is in the 
> need to convert Windows paths to UNIX ones.  R2WinBUGS does this using 
> undocumented and unsupported features of some versions of WINE.  As is 
> the way with unsupported features these have been changed on recent 
> versions of WINE and broken R2WinBUGS.
> The fix is not trivial, and the WINE support stuff in R2WinBUGS needs to 
> be completely rewritten.  I can see what needs to be done, and if I find 
> the time, I will fix it.

Contributions are welcome. I do not touch that xyz...-WINE conversion 
stuff at all.

Best wishes,

> You can also use JAGS ( http://www-fis.iarc.fr/~martyn/software/jags/ ) 
> which has now reached version 1.0.1 and has an R support package rjags.  
> I hope to post a fat Mac binary for this before the end of the week.  In 
> testing, this works on Leopard with a 32 bit R controlling a 64 bit JAGS 
> binary which is sort of cool.  Now all we need is to be able to run 
> multiple chains on different CPUs but there are some issues with the 
> RNGs from R which are used in JAGS.
> Bill Northcott

More information about the R-help mailing list