fillpattern
Pattern fills for ggplot2, hardened against the ways users write sizes
A side-by-side editorial comparison of detectseparation and haze — release velocity, themes, recent moves, and the top alternatives to consider.
A diagnostic package that generalized past its own name, then learned to say which kind of separation it found
detectseparation identifies separation and infinite estimates in binomial-response GLMs — the condition where maximum likelihood estimates diverge and standard software reports large coefficients with enormous standard errors instead of an error. Version 0.3 was the structural turn: detect_infinite_estimates() became the general method covering log, logit, probit and cauchit links, with detect_separation() demoted to a wrapper around it. Version 0.4 in April 2026 adds the ability to distinguish complete from quasi-complete separation via separation_type.
Four dormant years end with a modernization pass and an off-by-one fix in the C++ core
haze does nearest-neighbour smoothing and k-d tree interpolation on brain surface meshes. It sat untouched from April 2022 until July 2026, when a single release modernized it for current R versions and corrected an off-by-one error in the C++ code. It is not on CRAN and never will be — the package exceeds 50MB against CRAN's 5MB ceiling, a constraint its own initial release notes acknowledge.
detectseparation identifies separation and infinite estimates in binomial-response GLMs — the condition where maximum likelihood estimates diverge and standard software reports large coefficients with enormous standard errors instead of an error. Version 0.3 was the structural turn: detect_infinite_estimates() became the general method covering log, logit, probit and cauchit links, with detect_separation() demoted to a wrapper around it. Version 0.4 in April 2026 adds the ability to distinguish complete from quasi-complete separation via separation_type.
The package has been generalizing steadily — first past its own framing, since separation is one case of infinite estimates rather than the whole problem, and now toward finer classification of what it detects. The distinction 0.4 adds is practically useful because complete and quasi-complete separation call for different responses. Release intervals are long, roughly two to four years, which fits a diagnostic tool whose underlying theory is settled.
With link coverage broad and separation now classified by type, further work is more likely to refine reporting than to extend detection to new model families.
haze does nearest-neighbour smoothing and k-d tree interpolation on brain surface meshes. It sat untouched from April 2022 until July 2026, when a single release modernized it for current R versions and corrected an off-by-one error in the C++ code. It is not on CRAN and never will be — the package exceeds 50MB against CRAN's 5MB ceiling, a constraint its own initial release notes acknowledge.
The July 2026 release arrived 56 minutes after its sibling regfusionr 0.3.0 from the same maintainer, which is the tell: this is a maintainer sweeping a set of related neuroimaging packages back into working order, not independent development on haze itself. haze is the dependency, regfusionr the consumer, and the substantive work sits on the regfusionr side. The off-by-one correction is the only change here that alters results.
Expect haze to move only when a downstream dfsp-spirit package needs it to — its cadence is driven by the sibling packages, not by its own roadmap.
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 detectseparation or haze.
Pattern fills for ggplot2, hardened against the ways users write sizes
gcube's recent releases are all packaging metadata, not simulation code
The R port of Quinlan's Cubist gets reproducibility fixes, not new modelling
ggstats keeps widening what a coefficient or Likert plot can be
ecodive rebuilt itself into a broad diversity-metric library, breaking as it went
State-space data simulation for R, filled in one function at a time
See all detectseparation alternatives → · See all haze alternatives →
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
Both compete on the same themes — r-package — within Analytics. haze is currently shipping more aggressively (velocity 2.5 vs 0.0), with 0 editorial sparks in the last 30 days against 0. 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. haze is currently shipping more aggressively (velocity 2.5 vs 0.0), with 0 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other Analytics products to evaluate alongside.
Top detectseparation alternatives in Analytics are ranked by recent ship velocity. Browse the "detectseparation alternatives" section above for the current picks, or visit /alternatives/detectseparation for the full list with editorial commentary on each.
Top haze alternatives in Analytics are ranked by recent ship velocity. Browse the "haze alternatives" section above for the current picks, or visit /alternatives/haze for the full list with editorial commentary on each.