Applied researchers interested in Bayesian statistics are increasingly
attracted to R because of the ease of which one can code algorithms to sample
from posterior distributions as well as the significant number of packages
contributed to the Comprehensive R Archive Network (CRAN) that provide tools for Bayesian inference.
This task view catalogs these tools. In this task view, we divide those packages
into four groups based on the scope and focus of the packages. We first review R packages
that provide Bayesian estimation tools for a wide range of models. We then discuss
packages that address specific Bayesian models or specialized methods in Bayesian statistics.
This is followed by a description of packages used for postestimation analysis.
Finally, we review packages that link R to other Bayesian sampling engines such as
JAGS
,
OpenBUGS
, and
WinBUGS
.
Bayesian packages for general model fitting

The
arm
package contains R functions for Bayesian inference using lm, glm,
mer and polr objects.

BACCO
is an R bundle for Bayesian analysis of random functions.
BACCO
contains three subpackages: emulator, calibrator, and approximator,
that perform Bayesian emulation and calibration of computer programs.

bayesm
provides R functions for Bayesian inference for various models
widely used in marketing and microeconometrics. The models include linear regression models,
multinomial logit, multinomial probit, multivariate probit, multivariate mixture of
normals (including clustering), density estimation using finite mixtures of normals
as well as Dirichlet Process priors, hierarchical linear models, hierarchical multinomial logit,
hierarchical negative binomial regression models, and linear instrumental variable models.

bayesSurv
contains R functions to perform Bayesian inference for survival
regression models with flexible error and random effects distributions.

DPpackage
contains R functions for Bayesian nonparametric and semiparametric models.
DPpackage currently includes semiparametric models for density estimation, ROC curve analysis,
interval censored data, binary regression models, generalized linear mixed models, and IRT type models.

MCMCpack
provides modelspecific Markov chain Monte Carlo (MCMC) algorithms for
wide range of models commonly used in the social and behavioral sciences. It contains R functions
to fit a number of regression models (linear regression, logit, ordinal probit, probit, Poisson regression, etc.),
measurement models (item response theory and factor models), changepoint models (linear regression,
binary probit, ordinal probit, Poisson, panel), and models for ecological inference.
It also contains a generic Metropolis sampler that can be used to fit arbitrary models.

The
mcmc
package consists of an R function for a randomwalk Metropolis algorithm for
a continuous random vector.
Bayesian packages for specific models or methods

abc
package implements several ABC algorithms for performing parameter estimation and model selection.
Crossvalidation tools are also available for measuring the accuracy of ABC estimates, and to calculate the misclassification
probabilities of different models.

AdMit
provides functions to perform the fitting of an adapative mixture of Studentt distributions
to a tgarget density through its kernel function. The mixture approximation can be used as the importance density
in importance sampling or as the candidate density in the MetropolisHastings algorithm.

The
BaBooN
package contains two variants of Bayesian Bootstrap Predictive Mean Matching to
multiply impute missing data.

The
bark
package impelements BARK (Bayesian Additive Regression Kernels)
with feature selection.

The
BAS
package implements BMA for regression models using gpriors and mixtures of gpriors.
BAS
utilizes an efficient aglorithm to sample models without replacement.

The
bayesGARCH
package provides a function which perform the Bayesian estimation of the
GARCH(1,1) model with Student's t innovations.

Bayesthresh
fits a linear mixed model for ordinal categorical responses using
Bayesian inference via Monte Carlo Markov Chains. Default is Nandran and Chen algorithm
using Gaussian link function and saving just the summaries of the chains.

BayesTree
implements BART (Bayesian Additive Regression Trees)
by Chipman, George, and McCulloch (2006).

bayesQR
supports Bayesian quantile regression using the asymmetric Laplace distribution,
both continuous as well as binary dependent variables.

BayHaz
contains a suite of R functions for Bayesian estimation of smooth hazard rates via
Compound Poisson Process (CPP) priors.

BAYSTAR
provides functions for Bayesian estimation of threshold autoregressive models.

bbemkr
implements Bayesian bandwidth estimation for NadarayaWatson
type multivariate kernel regression with Gaussian error.

BCE
contains function to estimates taxonomic compositions from
biomarker data using a Bayesian approach.

BCBCSF
provides functions to predict the discrete response based on selected high dimensional features,
such as gene expression data.

bclust
builds a dendrogram with log posterior as a natural distance defined by the model.
It is also capable to computing Bayesian discrimination probabilities equivalent to the implemented Bayesian clustering.
SpikeandSlab models are adopted in a way to be able to produce an importance measure for clustering
and discriminant variables.

bcp
implements a Bayesian analysis of changepoint problem using
Barry and Hartigan product partition model.

bfp
implements the Bayesian paradigm for fractional polynomial models
under the assumption of normally distributed error terms.

bisoreg
implements the Bayesian isotonic regression with Bernstein polynomials.

BLR
provides R functions to fit parametric regression models using different types of shrinkage methods.

The
BMA
package has functions for Bayesian model averaging for linear models,
generalized linear models, and survival models. The complementary package
ensembleBMA
uses
the
BMA
package to create probabilistic forecasts of ensembles using a mixture of
normal distributions.

BMS
is Bayesian Model Averaging library for linear models with a wide choice of (customizable) priors.
Builtin priorss include coefficient priors (fixed, flexible and hyperg priors), and 5 kinds of model priors.

Bmix
is a barebones implementation of sampling algorithms for a variety of Bayesian stickbreaking
(marginally DP) mixture models, including particle learning and Gibbs sampling for static DP mixtures,
particle learning for dynamic BAR stickbreaking, and DP mixture regression.

bnlearn
is a package for Bayesian network structure learning
(via constraintbased, scorebased and hybrid algorithms), parameter learning (via ML and Bayesian estimators) and inference.

bqtl
can be used to fit quantitative trait loci (QTL) models.
This package allows Bayesian estimation of multigene models
via Laplace approximations and provides tools for interval mapping of genetic loci.
The package also contains graphical tools for QTL analysis.

bspec
performs Bayesian inference on the (discrete) power spectrum of time series.

bspmma
is a package for Bayesian semiparametric models for metaanalysis.

BSquare
models the quantile process as a function of predictors.

BVS
is a package for Bayesian variant selection and Bayesian model uncertainty
techniques for genetic association studies.

catnet
is a package that handles discrete Bayesian network models and provides
inference using the frequentist approach.

cslogistic
has a function that performs a Bayesian analaysis of
a conditionally specified logistic regression model.

dclone
provides low level functions for implementing maximum likelihood estimating procedures
for complex models using data cloning and MCMC methods.

deal
provides R functions for Bayesian network analysis;
the current version of covers discrete and continuous variables
under Gaussian network structure.

dlm
is a package for Bayesian (and likelihood) analysis of
dynamic linear models. It includes the calculations of
the Kalman filter and smoother, and the forward filtering backward sampling algorithm.

EbayesThresh
implements Bayesian estimation for thresholding methods.
Although the original model is developed in
the context of wavelets, this package is useful when researchers need to take
advantage of possible sparsity in a parameter set.

ebdbNet
can be used to infer the adjacency matrix of a network from time course data
using an empirical Bayes estimation procedure based on Dynamic Bayesian Networks.

eigenmodel
estimates the parameters of a model for symmetric relational data
(e.g., the abovediagonal part of a square matrix), using a modelbased eigenvalue decomposition and regression using MCMC methods.

evdbayes
provides tools for Bayesian analysis of extreme value models.

exactLoglinTest
provides functions for loglinear models that compute
Monte Carlo estimates of conditional Pvalues for goodness of fit tests.

factorQR
is a package to fit Bayesian quantile regression models
that assume a factor structure for at least part of the design matrix.

FME
provides functions to help in fitting models to data, to perform Monte Carlo,
sensitivity and identifiability analysis. It is intended to work with models be written as a set of
differential equations that are solved either by an integration routine from deSolve,
or a steadystate solver from rootSolve.

The
gbayes()
function in
Hmisc
derives the posterior
(and optionally) the predictive distribution when both the
prior and the likelihood are Gaussian, and when the statistic of interest comes from a twosample problem.

ggmcmc
is a tool for assessing and diagnosing convergence of Markov Chain Monte Carlo simulations,
as well as for graphically display results from full MCMC analysis.

growcurves
is a package for Bayesian semi and nonparametric growth curve models
that additionally include multiple membership random effects.

The
HI
package has functions to implement a geometric approach
to transdimensional MCMC methods and random direction multivariate
Adaptive Rejection Metropolis Sampling.

The
hbsae
package provides functions to compute small area estimates based on a basic area or unitlevel model.
The model is fit using restricted maximum likelihood, or in a hierarchical Bayesian way.

iterLap
performs an iterative Laplace approximation to build a global approximation of the posterior (using
mixture distributions) and then uses importance sampling for simulation based inference.

The function
krige.bayes()
in the
geoR
package performs
Bayesian analysis of geostatistical data allowing
specification of different levels of uncertainty in the model parameters.
The
binom.krige.bayes()
function in the
geoRglm
package implements Bayesian posterior simulation and spatial prediction
for the binomial spatial model (see the
Spatial
view for more information).

The
lmm
package contains R functions to fit linear mixed models using MCMC methods.

MasterBayes
is an R package that implements MCMC methods to
integrate over uncertainity in pedigree configurations estimated from molecular markers and phenotypic data.

MCMCglmm
is package for fitting Generalised Linear Mixed Models using MCMC methods.

The
mcmcsamp()
function in
lme4
allows MCMC sampling for
the linear mixed model and generalized linear mixed model.

The
mlogitBMA
Provides a modified function
bic.glm()
of the
BMA
package that
can be applied to multinomial logit (MNL) data.

The
MNP
package fits multinomial probit models using MCMC methods.

mombf
performs model selection based on nonlocal priors, including MOM, eMOM and iMOM priors..

monomvn
is an R package for estimation of multivariate normal and Studentt data of arbitrary
dimension where the pattern of missing data is monotone.

MSBVAR
is an R package for estimating Bayesian Vector
Autoregression models and Bayesian structural Vector Autoregression models.

pacbpred
perform estimation and prediction in highdimensional additive models,
using a sparse PACBayesian point of view and a MCMC algorithm.

PottsUtils
comprises several functions related to the Potts model definedon undirected graphs.

predmixcor
provides functions to predict the binary response based
on high dimensional binary features modeled with Bayesian mixture models.

prevalence
provides functions for the estimation of true prevalence from apparent prevalence in a
Bayesian framework. MCMC sampling is performed via JAGS/rjags.

profdpm
facilitates profile inference
(inference at the posterior mode) for a class of product
partition models.

The
pscl
package provides R functions to fit
itemresponse theory models using MCMC methods
and to compute highest density regions for the Beta distribution and the inverse gamma distribution.

The
PAWL
package implements parallel adaptive
MetropolisHastings and sequential Monte Carlo samplers for sampling from multimodal target distributions.

PReMiuM
is a package for profile regression, which is a Dirichlet process Bayesian clustering
where the response is linked nonparametrically to the covariate profile.

rcppbugs
is a package that attempts to provide an R alternative to using OpenBUGS/WinBUGS/JAGS
using random walk Metropolis sampling.

The
RJaCGH
package implements Bayesian analysis of
CGH microarrays using hidden Markov chain models.
The selection of the number of states is made via their posterior
probability computed by reversible jump Markov chain Monte Carlo Methods.

The
hitro.new()
function in
Runuran
provides an MCMC sampler
based on the HitandRun algorithm in combinaton with the RatioofUniforms method.

RSGHB
can be used to estimate models using a hierarchical Bayesian framework and provides flexibility in allowing
the user to specify the likelihood function directly instead of assuming predetermined model structures.

rstiefel
simulates random orthonormal matrices from linear and quadratic exponential family distributions on the
Stiefel manifold using the Gibbs sampling method. The most general type of distribution covered is
the matrixvariate Binghamvon MisesFisher distribution.

RxCEcolInf
fits the R x C inference model described in Greiner and Quinn (2009).

SampleSizeMeans
contains a set of R functions for calculating sample size requirements using
three different Bayesian criteria in the context of designing an
experiment to estimate a normal mean or the difference between two
normal means.

SampleSizeProportions
contains a set of R functions for calculating sample size requirements using
three different Bayesian criteria in the context of designing an
experiment to estimate the difference between two binomial
proportions.

sbgcop
estimates parameters of a Gaussian copula, treating the univariate marginal distributions as
nuisance parameters as described in Hoff(2007). It also provides a semiparametric imputation procedure for missing
multivariate data.

SimpleTable
provides a series of methods to conduct Bayesian inference and sensitivity analysis
for causal effects from 2 x 2 and 2 x 2 x K tables.

sna, an R package for social network analysis, contains functions to generate posterior samples from Butt's
Bayesian network accuracy model using Gibbs sampling.

spBayes
provides R functions that fit Gaussian spatial process models
for univariate as well as multivariate pointreferenced data using MCMC methods.

spikeslab
provides functions for prediction and variable selection using spike and slab regression.

spikeSlabGAM
implements Bayesian variable selection, model choice, and regularized estimation in
(geo)additive mixed models for Gaussian, binomial, and Poisson responses.

spTimer
fits, spatially predict and temporally forecast large amounts of spacetime data using
Bayesian Gaussian Process Models, Bayesian AutoRegressive (AR) Models, and Bayesian Gaussian Predictive Processes
based AR Models.

stochvol
provides efficient algorithms for fully Bayesian estimation of stochastic volatility (SV) models.

The
tgp
package implements Bayesian treed Gaussian process models:
a spaptial modeling and regression package providing fully Bayesian MCMC posterior inference for
models ranging from the simple linear model, to nonstationary treed Gaussian process, and others in between.

vbmp
is a package for variational Bayesian multinomial probate regression with Gaussian process priors.
It estimates class membership posterior probability employing variational and sparse approximation to the full posterior.
This software also incorporates feature weighting by means of Automatic Relevance Determination.

The
varSelectIP
package implements objective Bayes variable selection in linear regression and probit models.

The
vcov.gam()
function the
mgcv
package can extract a
Bayesian posterior covariance matrix of the parameters from a fitted
gam
object.

zic
provides functions for an MCMC analysis of zeroinflated count models including stochastic search variable selection.
Postestimation tools

BayesValidate
implements a software validation method for Bayesian softwares.

The
boa
package provides functions for diagnostics,
summarization, and visualization of MCMC sequences. It imports
draws from BUGS format, or from plain matrices.
boa
provides the Gelman and Rubin,
Geweke, Heidelberger and Welch, and Raftery and Lewis diagnostics,
the Brooks and Gelman multivariate shrink factors.

The
coda
(Convergence Diagnosis and Output Analysis) package is a suite of functions that
can be used to summarize, plot, and and diagnose convergence from MCMC samples.
coda
also defines an
mcmc
object
and related methods which are used by other packages.
It can easily import MCMC output from WinBUGS, OpenBUGS, and JAGS, or from plain
matrices.
coda
contains the Gelman and Rubin, Geweke, Heidelberger and Welch, and Raftery and Lewis diagnostics.

ramps
implements Bayesian geostatistical analysis of Gaussian processes
using a reparameterized and marginalized posterior sampling algorithm.
Packages for learning Bayesian statistics

AtelieR
is a GTK interface for teaching basic concepts in statistical inference,
and doing elementary bayesian statistics (inference on proportions, multinomial counts, means and variances).

The
BaM
package is an R package associated with Jeff Gill's book,
"Bayesian Methods: A Social and Behavioral Sciences Approach, Second Edition" (CRC Press, 2007).

BayesDA
provides R functions and datasets for "Bayesian Data Analysis, Second Edition" (CRC Press, 2003)
by Andrew Gelman, John B. Carlin, Hal S. Stern, and Donald B. Rubin.

The
Bolstad
package contains a set of R functions and data sets for the book
Introduction to Bayesian Statistics, by Bolstad, W.M. (2007).

The
LearnBayes
package contains a collection of functions helpful
in learning the basic tenets of Bayesian statistical inference. It contains functions for summarizing basic one and two
parameter posterior distributions and predictive distributions and MCMC algorithms for
summarizing posterior distributions defined by the user. It also contains functions for regression
models, hierarchical models, Bayesian tests, and illustrations of Gibbs sampling.
Packages that link R to other sampling engines

bayesmix
is an R package to fit Bayesian mixture models
using
JAGS
.

BayesX
provides functionality for exploring and visualizing estimation results
obtained with the software package
BayesX
.

BRugs
provides an R interface to
OpenBUGS
.
It works under Windows and Linux.
BRugs
used to be available from CRAN, now it is
located at the
CRANextras
repository.

cudaBayesreg
provides a Compute Unified Device Architecture (CUDA) implementation of a Bayesian multilevel model
for the analysis of brain fMRI data.
CUDA is a software platform for massively parallel highperformance computing on NVIDIA GPUs.

There are two packages that can be used to interface R with
WinBUGS
.
R2WinBUGS
provides a set of functions to call WinBUGS on
a Windows system and a Linux system;
rbugs
supports Linux systems through
OpenBUGS
on Linux (LinBUGS).

glmmBUGS
writes BUGS model files, formats data, and creates starting values for generalized linear mixed models.

There are three packages that provide R interface with
Just Another Gibbs Sampler (JAGS)
:
rjags,
R2jags, and
runjags.

All of these BUGS engines use graphical models for model specification. As such, the
gR
task view may be of interest.

Note that
rcppbugs
is a package that attempts to provide a pure R alternative to using OpenBUGS/WinBUGS/JAGS for MCMC.
The Bayesian Inference Task View is written by Jong Hee Park (Seoul National University, South Korea),
Andrew D. Martin (Washington University, St. Louis, MO, USA),
and Kevin M. Quinn (UC Berkeley, Berkeley, CA, USA).
Please email the
task view maintainer
with suggestions.