daniel1noble / BACE / 25329550598
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
25329550598

push

github

daniel1noble 
Pigauto alignment phase 1: mean_baseline + long-format output + lift_pct

Brings the BACE benchmark output into structural alignment with
pigauto's bench scripts (https://github.com/itchyshin/pigauto)
for head-to-head comparison. Source-of-truth for the alignment:
script/bench_<dataset>.R skeleton + script/bench_phase_c_cross
_dataset.R aggregation pattern.

Engine changes (dev/benchmark_engine.R):

  1. New helper .compute_baseline_metrics(). Predicts column-mean
     for continuous/count traits and modal class for factor types.
     Computes the same metric set as the bace path (rmse / nrmse /
     mae_fit / mae_raw / accuracy / balanced_accuracy / brier /
     mae_level) so the two methods stack cleanly. correlation +
     coverage95 are NA for the constant predictor (correctly --
     pigauto reports them as NA too).

  2. Each metric row tagged with method = "bace" or
     method = "mean_baseline". Wide summary_metrics.csv now has
     2 rows per trait (one per method). Backward-compat: legacy
     single-method bundles are read with method coerced to "bace".

  3. New summary_metrics_long.csv emitted alongside the wide form:
     pigauto-style tidy (dataset, method, trait, type, scale,
     n_hidden, metric, value). One row per (method, trait, metric).
     NA values filtered out. Easier to stack across runs / methods
     for cross-suite comparison.

  4. New rmse column (raw, on fit scale) alongside the existing
     nrmse. Pigauto's headline is raw rmse; nrmse stays for
     scale-free comparability.

  5. n_final stays at 10 for cloud feasibility (pigauto canonical
     N_IMP=20 doubles final-imputation wall time, pushes BACE past
     the GHA 5h45m budget on heavier datasets). Documented in the
     parameter comment so local-runs aiming for direct pigauto
     comparison know to override.

Report changes (dev/make_benchmark_report.R):

  6. New "BACE vs mean-baseline head-to-head" section with two
     tables: continuous/count traits showi... (continued)

1574 of 3667 relevant lines covered (42.92%)

39.49 hits per line

Jobs
ID Job ID Ran Files Coverage
1 25329550598.1 15
42.92
 GitHub Action Run
Source Files on build 25329550598