LoginSign Up
← Back to all sparks
B

bayestools

ANALYTICS
Velocity0.0

R package bayestools by fbartos — release notes from GitHub.

github.com/fbartos/bayestools ↗

The JAGS toolkit under RoBMA, shipping the standardization machinery its downstream rewrite needed

r-packagebayesianjagspriorsinfrastructure
◆Current state
BayesTools provides the shared JAGS fitting, prior and summary infrastructure that the author's meta-analysis packages build on. The 0.2.x line filled in modeling primitives — prior_mixture() and mixed-posterior objects in 0.2.18, expression-valued priors and lme4-style uncorrelated random effects in 0.2.20, then a run of small diagnostic fixes for mixture and spike-and-slab priors. Version 0.3.0 in May 2026 adds automatic standardization of continuous predictors, default priors for unspecified factor and continuous terms, and functions to transform prior and posterior samples back to the original scale.
◆Where it's heading
This package's releases are best read against what depends on them. The 0.2.x fixes track features appearing in RoBMA one version later, and 0.3.0 landed a single day before RoBMA 4.0.0 — the standardization and sample-transformation functions are the substrate that rewrite needed. The direction of the work is toward sensible defaults: default priors by predictor type, automatic standardization for sampling stability, and transformation back to interpretable scale so the convenience does not cost the user their units.
◆Prediction
Given how tightly its releases track downstream needs, the next version is most likely driven by gaps surfacing in RoBMA 4.0.x rather than by independent feature work.

◆Recent moves

  1. 3mo ago

    Adds automatic predictor standardization and type-based default priors

    Adds a formula_scale parameter that standardizes continuous predictors for sampling stability, default priors for unspecified factor and continuous terms, and transform functions returning prior and posterior samples to the original scale. It shipped one day before RoBMA 4.0.0, whose rewrite depends on this machinery — the origin release of a two-package wave.

    View source ↗
  2. 8mo ago

    BayesTools 0.2.23

    Fixes JAGS_diagnostics handling of factor parameters nested inside mixture priors. A narrow correction to the mixture-prior support added in 0.2.18.

    View source ↗
  3. 8mo ago

    BayesTools 0.2.22

    Fixes plot_posterior() for spike and slab priors. Released minutes before 0.2.23 and part of the same cleanup pass over spike-and-slab and mixture plotting.

    View source ↗
  4. 11mo ago

    BayesTools 0.2.21

    Restores an intercept placeholder in JAGS_formula() so the model matrix keeps its shape, and repairs the silent argument on JAGS_fit(). Small correctness fixes in the formula-generation path.

    View source ↗
  5. 1y ago

    BayesTools 0.2.20

    Lets prior functions accept expressions so priors can depend on other JAGS parameters, allows literal text appended to generated formulas, and adds lme4-style uncorrelated random effects via (x||y). This is the release that turns the formula interface into something users can extend rather than only configure.

    View source ↗
  6. 1y ago

    BayesTools 0.2.19

    Adds a max_extend cap on autofit iterations and JASP progress bar integration, plus fixes across mixture diagnostics and formula evaluation. The JASP hook is a reminder that this infrastructure also serves a point-and-click statistics application, not only R callers.

    View source ↗