Bug#169717: ITP: r-cran-mcmcpack -- routines for Markov Chain Monte Carlo model estimation in R
Package: wnpp
Version: unavailable; reported 2002-11-19
Severity: wishlist
* Package name : r-cran-mcmcpack
Version : 0.1.4
Upstream Authors: Andrew D. Martin <admartin@artsci.wustl.edu>,
Kevin M. Quinn <quinn@stat.washington.edu>
* URL : http://scythe.wustl.edu/mcmcpack.html
* License : GPL
Description : routines for Markov Chain Monte Carlo model estimation in R
This is a set of routines for R that implement various statistical
models using Markov Chain Monte Carlo estimation with Gibbs sampling,
which allows "solving" models that would otherwise be intractable with
traditional techniques, particularly problems in Bayesian statistics
(where one or more "priors" are used as part of the estimation
procedure, instead of an assumption of ignorance about the "true"
point estimates), although MCMC can also be used to solve frequentist
statistical problems without priors.
Currently implemented are a number of ecological inference routines
(for estimating individual attributes from aggregate data, such as
electoral returns or census results), as well as models for
traditional linear panel and cross-sectional data and some
visualization routines for diagnostics. Additional models, including
more EI routines (such as the model developed by Gary King), an
item-response theory model, and a binary response (probit) model are
planned for future releases.
(Description from the upstream website.)
MCMCpack is an R package that employs Markov chain Monte Carlo (MCMC)
methods to fit commonly used statistical models. R is an extremely
powerful language and environment for statistical computation and
graphics. We presume that users of this package will be familiar with
R.
Currently MCMCpack allows the user to simulate from the posterior
density of the following models: linear regression (with Gaussian
errors), a general linear panel model, Wakefield's ecological
inference model, Quinn's dynamic ecological inference model, and
Wakefield's hierarchial ecological inference model. Soon we will make
available code to estimate a probit model, and a one-dimensional item
response theory model. The package also contains densities and random
number generators for commonly used distributions that are not part of
the standard R distribution, some additional functions that are useful
for manipulating mcmc objects, and some data visualization tools for
ecological inference.
(Personal aside: this is one *cool* cutting-edge package, even if only
about six of us Debianistas will appreciate it. And it's co-written
by a political scientist to boot. :-)
I will also package the Scythe C++ library and the coda package from
CRAN [even though it's pure R code] that are required for this package
to work. Both are also GPLed.)
-- System Information:
Debian Release: testing/unstable
Architecture: i386
Kernel: Linux quantex.lordsutch.com 2.4.20-pre10-ac1 #2 Wed Oct 30 15:27:03 CST 2002 i686
Locale: LANG=en_US, LC_CTYPE=en_US
Reply to: