Social Science Statistics Blog

I don't know anything about it, but it seems there are a few people at Nuffield who are interested in this sort of thing: http://www.nuff.ox.ac.uk/Economics/Papers/2004/w16/JADDFHNSHandbook.pdf

Thanks for the link Tim, this looks very interesting. I also just got a response from the techical System's Engineer's at SAS on the issue of multithreading. They say that:

"When you install SAS9.1.3 (our latest release) it comes with the threads option turned on by default and the cpucount option listed as 1. If you want to take advantage of the cpu's you will need to change the cpucount option to the number of cpu's in the machine. This can be done in the SAS System Options, the configuration file, within an option statement, or SAS invocation, see the below link for the online documentation (at the bottom is additional info on the cpucount option).
http://support.sas.com/onlinedoc/913/getDoc/en/lrdict.hlp/a002197022.htm

Not all SAS procedures are currently multi-threaded, we continue to add new ones, in the link above it lists what BASE SAS procedures and SAS STAT procedures are multi-threaded.

There is additional overhead associated with running the multi-threads options, I have seen very significant gains running multi-threaded programs on 4 processors and above, not as much on a two way box, and have not seen any numbers on a PC Operating System."

There is another solution for R at LindaSpaces, called NetWorkSpaces. The documentation makes it seem a bit simpler than SNOW.

I just stumbled across SciPy, and found that the developers there are working on this issue as well. See http://svn.scipy.org/svn/scipy/trunk/Lib/sandbox/cow/cow.py. It looks like it's only in the testing sandbox at the moment, not in a release.

Disclaimer: I haven't tried SciPy yet, but as someone who finds that R and Stata to be almost impossible to make sense of, I can't wait to get back to the sane, truly object-oriented, world of Python. Hopefully SciPy offers enough features for me...