daniel1noble / BACE / 25324297612
91%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 15
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 42.923%. Remained the same
25324297612

push

github

daniel1noble 
Realign cross-dataset benchmark to Shinichi's pigauto spec (2026-05-04)

Bring the BACE cross-dataset cloud benchmark into alignment with
Shinichi's pigauto bench so the two suites can be compared headline-
to-headline. Spec source: pigauto bench summary 2026-05-04.

Per-dataset trait sets (now matching pigauto headline):
  avonet    : 4 cont (Mass, Beak.Length_Culmen, Tarsus.Length,
              Wing.Length) + 2 cat (Trophic.Level K=4,
              Primary.Lifestyle) + 1 ord (Migration K=3)
              -- was 10 mixed
  pantheria : 4 cont + 1 count (litter_size) + 2 ord
              (diet_breadth K=5, habitat_breadth K=3) + 1 binary
              (terrestriality)  -- already matched, no wrapper change
  amphibio  : 2 cont only (Body_size_mm, Body_mass_g) -- was skipped;
              re-enabled. Discrete columns (diurnal/nocturnal binary,
              diet_breadth ordinal, habitat categorical) intentionally
              skipped per pigauto v1 (Rphylopars singular on
              AmphiBIO taxonomic tree; AmphiBIO binaries are also
              presence-only which makes MCMCglmm threshold fitting
              fail with singular MM).
  bien      : 5 cont (height_m, leaf_area, sla, seed_mass,
              wood_density) -- already matched, no wrapper change

Datasets dropped from the cloud matrix (kept as wrappers in the repo
for local use): globtherm, leptraits. Both not in pigauto's bench.

Engine miss_frac default 0.10 -> 0.30 to match pigauto's MCAR rate.
This stresses the imputation pipeline more and produces metrics
directly comparable to the pigauto headline numbers.

Caveats noted in the AVONET wrapper:
  - Shinichi's pigauto bench reports Primary.Lifestyle K=7 levels;
    avonet_traits.rda built from this repo's AVONET.csv has K=5.
    Different AVONET version. Not a bug; flag for follow-up.
  - Shinichi's per-dataset n (post tree alignment) values are
    smaller than this repo's .rda counts -- pigauto uses curated
    phyloTraitData:... (continued)

1574 of 3667 relevant lines covered (42.92%)

39.49 hits per line

Jobs
ID Job ID Ran Files Coverage
1 25324297612.1 15
42.92
 GitHub Action Run
Source Files on build 25324297612