DynCount: Bayesian Dynamic Models for Count Time Series

Fits Bayesian state-space models for count time series using a latent log-rate (Poisson), latent logit (binomial) or latent additive-log-ratio (multinomial choice counts) formulation. Each latent trajectory follows a first-order random walk or a stationary AR(1) process and is sampled by Metropolis-within-Gibbs using the implied Gaussian Markov random field full conditionals. The latent increments can be Gaussian, Student-t, a finite scale mixture of normals, or follow a stochastic volatility process, and the Poisson and binomial families support zero inflation. It implements and extends the methodology of Zens and Bijak (2026) <doi:10.1214/26-AOAS2171>.

Version:	0.2.0
Depends:	R (≥ 3.5.0)
Imports:	generics, stats, graphics, grDevices, utils
Suggests:	coda, stochvol (≥ 3.0.2), testthat (≥ 3.0.0), knitr, rmarkdown
Published:	2026-09-28
DOI:	10.32614/CRAN.package.DynCount
Author:	Gregor Zens [aut, cre]
Maintainer:	Gregor Zens <zens at iiasa.ac.at>
License:	MIT + file LICENSE
NeedsCompilation:	no
Language:	en-GB
Citation:	DynCount citation info
Materials:	README, NEWS
CRAN checks:	DynCount results

Documentation:

Reference manual:	DynCount.html , DynCount.pdf
Vignettes:	Dynamic Models for Poisson, Binomial and Multinomial Time Series (source, R code)

Downloads:

Package source:	DynCount_0.2.0.tar.gz
Windows binaries:	r-devel: DynCount_0.2.0.zip, r-release: DynCount_0.2.0.zip, r-oldrel: DynCount_0.2.0.zip
macOS binaries:	r-release (arm64): DynCount_0.2.0.tgz, r-oldrel (arm64): DynCount_0.2.0.tgz, r-release (x86_64): DynCount_0.2.0.tgz, r-oldrel (x86_64): DynCount_0.2.0.tgz
Old sources:	DynCount archive

Linking:

Please use the canonical form https://CRAN.R-project.org/package=DynCount to link to this page.