I stumbled on the shotgun library from reading Dr. Bickson's (somehow)
recent blog post:
I thought it could come in handy on some of the stuff I'm playing with
and wrote a wrapper for it so us R folks can use it, too ... I mean,
we can't lab the MATLAB tribe have all the fun, now, can we?
Some installation and usage instructions are on the wiki:
R has to be running in 64 bit for you to be able to build and install
the package successfully. It works on my OS X boxes, and our (linux)
compute server, so hopefully it can work for you.
It's a bit rough around the edges (ie. no documentation), but if
you're familiar with building models in R, you'll now how to use this:
R> model <- buckshot(A.arcene, y.arcene, 'logistic', lambda=0.166)
R> preds <- predict(model, A.arcene)
R> sum(preds == y.arcene) / length(y.arcene)
Could get worse than 100% accuracy, I guess ...
In time, I hope to get it "easily installable" and push it out to
CRAN, but in the meantime I thought it would be of interest to the
people reading this list in its current form, and to the shotgun
authors (who I'm hoping are also reading this list), even if they
don't use R :-)
Thanks for putting shotgun out in the wild for us to tinker with!
Graduate Student: Computational Systems Biology
| Memorial Sloan-Kettering Cancer Center
| Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/
Thanks a lot Steve! We really appreciate your efforts. The shotgun code has been significantly improved over the last two weeks. We are looking for more users to beta test it on real data. Write me if you are trying our code!