LoginSign Up
← Back to all sparks
V

vegan

ANALYTICS
Velocity0.0

Community Ecology Package

github.com/vegandevs/vegan ↗

Community ecology's standard toolkit is retiring the functions a generation of scripts was built on.

community-ecologyordinationdeprecationpermutation-testsstatistical-software
◆Current state
vegan is the reference R package for ordination and diversity analysis in community ecology. The 2.7 line redesigned constrained ordination graphics from scratch, made adonis defunct in favor of adonis2, and moved permutation tests for partial RDA and db-RDA onto residualized predictors. The current 2.7-5 release is stabilization — parallel processing on Windows, ggvegan compatibility for tidy scores, and arrow label placement.
◆Where it's heading
Two long deprecation programs are converging. Plotting is being pushed out of the package: lattice functions are deprecated in favor of ggplot2 equivalents now living in ggvegan, a separate CRAN package, while vegan's own base graphics were rebuilt around independently configurable score types and pipe-based construction. Meanwhile the statistical core is being made internally consistent — add1 and drop1 no longer use different internal models, and summary is no longer an accepted route to ordination scores.
◆Prediction
Expect the remaining lattice functions to go defunct and more plotting responsibility to shift to ggvegan, leaving vegan as the statistical engine with a thin base graphics layer.

◆Recent moves

  1. 2mo ago

    Parallel processing stabilized, especially on Windows

    Improves parallel processing stability and fixes inconsistencies concentrated on Windows, along with better ggvegan compatibility for scores(tidy = TRUE) and arrow label placement under optimize = TRUE. Consolidation work on the 2.7 redesign rather than new capability.

    View source ↗
  2. 5mo ago

    Lattice plotting deprecated as ggvegan reaches CRAN

    Deprecates vegan's lattice functions now that ggplot2 equivalents exist in ggvegan, newly on CRAN. The clearest signal yet that plotting is being moved out of the package rather than modernized inside it.

    View source ↗
  3. 10mo ago

    Partial RDA permutation tests adopt residualized predictors

    Brings partial RDA and db-RDA permutation tests in line with the pattern partial CCA adopted in 2.6-6, and redesigns add1 and anova(by = 'axis'), which had internally used a partial model that no longer makes sense under residualized permutation. add1 and drop1 are now consistent for ordination models where earlier releases used different internal models. Also makes tibbles without rownames behave like data frames.

    View source ↗
  4. 1y ago

    Ordination graphics rebuilt and adonis made defunct

    ⚡ SPARK

    The pivot of the 2.7 line: constrained ordination graphics are redesigned around independently configurable score types and pipe-based construction, and adonis is finally removed after three years of escalating warnings. Everything since has been stabilizing this release.

    View source ↗
  5. 1y ago

    Release chasing undefined behavior in a C function

    Cut mainly to clear undefined behavior in a C function flagged by CRAN checks, with the maintainer candidly noting they could not replicate the failure and could not be certain of the fix. User-visible changes are few.

    View source ↗
  6. 1y ago

    pca, ca and pco added for unconstrained ordination

    Adds pca, ca and pco as direct entry points for unconstrained ordination, giving the common methods plain names instead of requiring them to be reached through more general functions. Mostly minor fixes otherwise.

    View source ↗