Skip to content

fix(summary): drop share_incl and the indicator transition counts from the summary tables; make extract_ess(estimator = "mixt") defunct - #224

Merged
MaartenMarsman merged 3 commits into
developfrom
fix/drop-share-incl
Nov 6, 2026
Merged

MaartenMarsman merged 3 commits into
developfrom
fix/drop-share-incl

Conversation

@MaartenMarsman

Copy link
Copy Markdown
Collaborator

Three removals from the summary tables and one lifecycle step. Follows #223, which took the transition counts out of the printed inclusion table only.

Summary tables

  • The pairwise table of a bgm() fit and the difference tables of a bgmCompare() fit lose the share_incl column, the inclusion part's fraction of the composite Monte Carlo variance of a weight. Nothing in the package read it; n_eff, the composite effective sample size it was computed beside, is unchanged.
  • The inclusion tables of both fits lose the directional transition counts n0->1 and n1->0, from the summary object as well as the print.
  • The co-clustering table under the stochastic block prior (summary(fit)$allocations) loses n0->0, n0->1, n1->0, n1->1 and the transition-based n_eff_mixt. The counts entered this table and the inclusion table in the same commit (2025-08-01).

Every table now holds mean, mcse, sd, n_eff (where defined) and Rhat. The one-sentence note from #222 on blank cells is unchanged.

extract_ess()

extract_ess(estimator = "mixt"), deprecated with a warning in 0.2.0.0, is defunct: the call stops through lifecycle::deprecate_stop() with a message naming the change and estimator = "rb". The argument stays in the signature for this release so the error is the lifecycle one; removing the argument is a later step. The transition ESS is no longer recomputed from the raw indicator draws, so the helper that did so is gone. Fits saved by bgms < 0.2.0.0 previously fell back to the transition ESS silently; their indicator element now errors with the same message extract_posterior_inclusion_probabilities() gives.

No external caller passes estimator: easybgm, the tutorial project and the budget scripts all call extract_ess(fit).

Documentation

The diagnostics vignette no longer describes the transition counts (the R-hat paragraph and one bullet). The roxygen for extract_ess(), extract_rhat() and extract_posterior_inclusion_probabilities() drops the transition estimator and the counts. NEWS: two bullets under Diagnostics, one under Deprecations.

Not changed

The C++ kernel .compute_indicator_ess_cpp still returns the four transition counts and n_eff_mixt; only its own unit tests read them now.

Verification

  • tests/testthat/test-rb-inclusion-probabilities.R 58, test-mcmc-diagnostics.R 171, test-extractor-functions.R 691, test-methods.R 1912, test-bgmCompare.R 96, test-summary-marking.R 38, test-regressions-2.R 34, test-sample-sbm-prior.R 13, test-sbm-correction.R 36; all pass. Snapshot summary-marking.md updated for the pairwise table.
  • R CMD check --as-cran: 0 errors, 0 warnings, 0 notes.
  • styler and lintr clean; man pages regenerated.
  • Behavioural comparison develop (4acc315) vs branch: dev/review-2026-08/repro/dropcols-demos.R and dropcols-compare.R (run by run-dropcols.sh), output in ~/bgms-review/dropcols-comparison.txt. It shows the column sets of every table before and after, the printed header rows, and the warning-to-error change of extract_ess(fit, estimator = "mixt") with both messages.

…m the summary tables; make extract_ess(estimator = "mixt") defunct

The pairwise table of a bgm() fit and the difference tables of a
bgmCompare() fit lose the share_incl column. The inclusion tables and the
stochastic-block co-clustering table lose the directional transition
counts n0->1 and n1->0; the co-clustering table also loses n0->0, n1->1 and
the transition-based n_eff_mixt. The tables hold mean, mcse, sd, n_eff and
Rhat.

extract_ess() keeps its estimator argument; "mixt", deprecated in 0.2.0.0,
now errors through lifecycle::deprecate_stop(). The transition ESS is no
longer recomputed from the raw indicator draws.

The diagnostics vignette no longer describes the transition counts, and
the roxygen for extract_ess(), extract_rhat() and
extract_posterior_inclusion_probabilities() drops the transition
estimator.
@MaartenMarsman
MaartenMarsman merged commit 1a123c9 into develop Nov 6, 2026
0 of 8 checks passed
@MaartenMarsman
MaartenMarsman deleted the fix/drop-share-incl branch November 6, 2026 09:59
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant