churon
churon is spending its entire release history getting a Rust ONNX binding through CRAN.
A side-by-side editorial comparison of cubist and robma — release velocity, themes, recent moves, and the top alternatives to consider.
The R port of Quinlan's Cubist gets reproducibility fixes, not new modelling
Cubist is the R interface to Quinlan's rule-based regression model, wrapping the original C sources behind an R API and feeding the tidymodels rules package. The 0.6.0 release adds a strip_time_stamps control that removes date, time and duration information from model output, and now errors rather than silently misbehaving when a date or date-time column is passed. Error reporting moves from base stop() and warning() to cli.
RoBMA 4.0 tears out its own constructor surface and rebuilds on one class hierarchy
RoBMA fits robust Bayesian model-averaged meta-analyses that adjust for publication bias. The 3.x line grew by accretion: separate constructors for each model family (RoBMA.reg, NoBMA, BiBMA and their .reg variants), a spike-and-slab algorithm in 3.3.0 that made estimation fast enough to matter, then a steady stream of post-estimation tooling gated on that algorithm — heterogeneity summaries, residuals, funnel plots, z-curve conversion, predict, extract, pooled and adjusted effects. Version 4.0.0 in May 2026 collapses all of it into a unified brma class hierarchy.
Cubist is the R interface to Quinlan's rule-based regression model, wrapping the original C sources behind an R API and feeding the tidymodels rules package. The 0.6.0 release adds a strip_time_stamps control that removes date, time and duration information from model output, and now errors rather than silently misbehaving when a date or date-time column is passed. Error reporting moves from base stop() and warning() to cli.
The direction is custodial: this is a mature algorithm with a stable definition, so the work is making a decades-old C codebase behave predictably inside a modern R workflow. The reproducibility thread is the clearest one — embedded timestamps mean two identical models compare as different objects, which breaks caching, testing and any workflow that hashes results. Alongside it runs slow C hygiene, from keyword symbol overwrites in 0.5.0 to unused-variable warnings in 0.6.0.
Expect continued small maintenance releases tracking CRAN compiler requirements and the needs of the rules package, with no change to the modelling algorithm itself.
RoBMA fits robust Bayesian model-averaged meta-analyses that adjust for publication bias. The 3.x line grew by accretion: separate constructors for each model family (RoBMA.reg, NoBMA, BiBMA and their .reg variants), a spike-and-slab algorithm in 3.3.0 that made estimation fast enough to matter, then a steady stream of post-estimation tooling gated on that algorithm — heterogeneity summaries, residuals, funnel plots, z-curve conversion, predict, extract, pooled and adjusted effects. Version 4.0.0 in May 2026 collapses all of it into a unified brma class hierarchy.
The 3.x series solved the modeling problem and left an interface problem behind: a caller had to know which of six constructors matched their data type, and argument names differed across them. 4.0.0 resolves that by making the model family a set of arguments rather than a function name, and by standardizing input naming on metafor-style conventions. It shipped one day after BayesTools 0.3.0, the author's own upstream infrastructure package, whose new standardization and prior-transformation machinery this rewrite depends on.
A rewrite this wide usually needs a follow-up, so expect 4.0.x patches addressing migration gaps as users hit the removed constructors and renamed arguments.
Other Analytics products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either cubist or robma.
churon is spending its entire release history getting a Rust ONNX binding through CRAN.
firatheme woke up after four years and started fixing what ggplot2 changed underneath it.
bagyo reached CRAN as a Philippine tropical cyclone dataset, with its tags stamped out of order.
rainette rebuilt its Reinert clustering in 0.2.0, tuned it in 0.3.0, and has coasted since.
questionr's survey helpers are finished; recent releases only keep pace with R-devel.
washdata is a fixed survey dataset; eight years of releases have changed only its packaging.
See all cubist alternatives → · See all robma alternatives →
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
Both compete on the same themes — r-package — within Analytics. cubist and robma are shipping at a similar cadence (velocity 0.0 vs 0.0, both within Sparkpulse's "active" band). See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. cubist and robma are shipping at a similar cadence (velocity 0.0 vs 0.0, both within Sparkpulse's "active" band). For your specific use case, the alternatives sections above list other Analytics products to evaluate alongside.
Top cubist alternatives in Analytics are ranked by recent ship velocity. Browse the "cubist alternatives" section above for the current picks, or visit /alternatives/cubist for the full list with editorial commentary on each.
Top robma alternatives in Analytics are ranked by recent ship velocity. Browse the "robma alternatives" section above for the current picks, or visit /alternatives/robma for the full list with editorial commentary on each.