pscl: classes and methods for R developed in the Political Science Computational Library
Monday July 10, 2006
I just got an update of my/our R package, pscl, out to the CRAN servers (presuming it passes their checks, passed checks on my box, but you never know). I’ve also posted a copy of the tar.gz file at the pscl site.
Some highlights:
- rather nifty class for roll call data, with print and summary methods that do reasonable things
- a function for reading roll call data directly from Keith Poole’s voteview site
- a way for subseting rollcall objects to get rid of unanimous roll calls, lop-sided roll calls, legislators with few votes
- user-specified encodings of roll call data (i.e., you can specify what codes are “Yea” votes, what are “Nay”), and that encoding is an attribute of the roll call object that is looked at by summary methods, model fitting functions etc
- our ideal function for fitting item-response models to roll call data via Markov chain Monte Carlo, with print, summary, predict, plot and graphical diagnostics of various kinds
- a few little Bayesian widgets I’ve written in the course of working on my book manuscript (e.g., how do you compute a highest density region for an inverse-Gamma density?)
- maximum likelihood analysis of zero-inflated and hurdle model for count data that I wrote for one of my classes here a couple of years back; again, these behave a lot like lm/glm model objects in R, so you can the usual things like summary, predict, etc with them.