Installer jags avec Homebrew, apt, MacPorts, Nix

Liens du projet

• Page d'accueil

Commandes et alias

• jags

Historique du projet

Martyn Plummer presented JAGS at the 2003 Distributed Statistical Computing workshop as a program for Bayesian graphical models using Gibbs sampling, aiming at compatibility with classic BUGS while leaving room for an R package interface. The project site later summarized the same design goals as a cross-platform BUGS engine, an extensible system for custom functions, distributions, and samplers, and a platform for Bayesian modelling experiments.

The SourceForge project became the authoritative home for source, manuals, and binaries. Its code tree identifies JAGS as a C++ implementation based on the BUGS program created by the MRC Biostatistics Unit, and the project site documents binary distribution for macOS and Windows plus separate Debian, Ubuntu, and MacPorts packaging.

Historique d'adoption

JAGS adoption is closely tied to the BUGS ecosystem and to R. The project site explicitly points users to Debian, Ubuntu, and MacPorts packages, while the rjags documentation describes an interface to the JAGS MCMC library maintained by Martyn Plummer. Homebrew API data available during this enrichment recorded 890 installs over 365 days, which is small compared with broad CLI tools but meaningful for a specialized scientific engine.

The package has persisted because it gives statisticians a scriptable, cross-platform BUGS-language engine without depending on the WinBUGS/OpenBUGS GUI tradition. That made it a practical backend for teaching, ecological models, Bayesian hierarchical workflows, and R-driven analysis pipelines.

Modes d'utilisation

Users write Bayesian hierarchical models in a BUGS-like language, run MCMC simulation through the `jags` terminal interface or an integration such as rjags, and post-process chains in statistical software. The package is a computation backend rather than a general-purpose interactive application.

Pourquoi les passionnés de paquets s'y intéressent

JAGS is package-nerd interesting because it is a long-lived SourceForge-era scientific tool that still matters at the package boundary: install the native sampler once, then R, scripts, courses, and reproducible analyses can target the same engine across Unix-like systems, Windows, and macOS.

Chronologie

• 2003: Plummer's DSC paper described JAGS as a BUGS-compatible Bayesian graphical modelling program.
• 2015: JAGS 4.0.0 NEWS recorded language and library changes including log-density functions, vector indexing, additional distributions, and better compiler diagnostics.
• 2022: JAGS 4.3.1 NEWS recorded a Windows installer update for the Rtools42 toolchain.
• 2023: The project site recorded JAGS 4.3.2 as released on March 4, 2023, and the JAGS News post described it as a patch release for evolving C++ standard requirements.
• 5.0.0: The NEWS file records new monitor types for deviance-related summaries and WAIC-oriented workflows.

Related projects

• JAGS belongs to the BUGS lineage with WinBUGS, OpenBUGS, and MultiBUGS, and it is commonly paired with R packages such as rjags. It is conceptually adjacent to later Bayesian engines such as Stan, but its historical niche is BUGS-language compatibility and Gibbs-sampling-oriented extensibility.

Sources

• https://cran.r-project.org/web/packages/rjags/rjags.pdf
• https://formulae.brew.sh/api/formula/jags.json
• https://martynplummer.wordpress.com/2023/03/04/jags-4-3-2-is-released/
• https://mcmc-jags.sourceforge.io/
• https://sourceforge.net/p/mcmc-jags/code-0/ci/default/tree/
• https://sourceforge.net/p/mcmc-jags/code-0/ci/default/tree/NEWS?format=raw
• https://www.r-project.org/conferences/DSC-2003/Proceedings/Plummer.pdf