gp3bayespy 0.6.2

What's Changed

• Synchronize documentation with gp3bayespy 0.6.1 by @stefanosbalaskas in #17
• Docs: illustrated estimation and validation figure gallery by @stefanosbalaskas in #20
• Research: October 2026 validation methods, regression tests, figures and articles by @stefanosbalaskas in #22
• Docs: showcase October scientific methods, figures and release boundaries by @stefanosbalaskas in #24
• Release candidate: gp3bayespy 0.6.2 research documentation maintenance by @stefanosbalaskas in #25

Full Changelog: v0.6.1...v0.6.2

gp3bayespy 0.6.1

gp3bayespy 0.6.1

Corrected contract-first Bayesian multilevel gaze-mediation release.

Certified gates

• PR #14 mediation tranche merged
• statement coverage: 100%
• branch-aware coverage: 100%
• missing statements: 0
• missing branches: 0
• coverage exclusions: 0
• Ruff: PASS
• mypy: PASS
• GitHub cross-platform CI: PASS
• Docs deployment: PASS
• all 13 examples: PASS
• exact wheel/sdist: PASS
• Twine: PASS
• clean wheel install: PASS
• clean sdist install: PASS

Release commit: 6f69aa5

The withdrawn v0.6.0 GitHub-only tag remains unchanged and was never published to PyPI.

The DOI 10.5281/zenodo.22150746 remains specific to gp3bayespy 0.5.0.

gp3bayespy 0.6.0 (withdrawn)

gp3bayespy 0.6.0 — withdrawn GitHub-only release

Do not use this release.

The v0.6.0 GitHub release workflow failed before PyPI publication.

The tagged commit did not contain the intended multilevel-mediation tranche from PR #14. The exact wheel therefore failed the required mediation API smoke test.

• gp3bayespy 0.6.0 was not published to PyPI.
• no validated 0.6.0 wheel or sdist was attached to this GitHub release.
• the v0.6.0 tag is intentionally retained unchanged as an audit record.
• the corrected release is gp3bayespy 0.6.1.

Failed tagged commit:

b9b06f0

gp3bayespy 0.5.0

gp3bayespy 0.5.0

First public Python release of gp3bayespy, frozen against gp3bayes 0.5.0.

Parity and validation

• 458/458 frozen R exports implemented.
• 59/59 canonical articles ported.
• 8/8 executable workflow examples.
• 0 public unrestricted **kwargs catchalls.
• 321/321 tests passing.
• Branch-aware coverage: 47.9482%.
• Ruff: PASS.
• mypy: PASS.
• Linux / macOS / Windows CI across Python 3.11, 3.12 and 3.13: PASS.
• .[all] CI: PASS.
• Strict MkDocs: PASS.
• Clean wheel and sdist with twine check: PASS.
• Fresh installed-wheel parity and examples: PASS.

Frozen provenance

• Frozen R reference: gp3bayes 0.5.0.
• Frozen R reference commit: 0ad32d8921c3b267d2021324c3b3568edf3ae01f.
• Validated parity source commit: d9fa3cdaa0ac261cd5b3cb39175d3896d9c8a1c7.
• Deep-freeze metadata commit: 791704a0897d3677b8c6d7eb34f8f0aa284a85db.
• Final Python release commit: 5840e7cb04640a1ea79a38cf91a7545e2a92d7ef.
• Deep-freeze record: dev/parity/final_deep_freeze_0.5.0.json.

Artifacts

• gp3bayespy-0.5.0-py3-none-any.whl
  • SHA-256: 56de34188c7146a7c60af510bf0b8ad346b9dde893686f73eae923a15a098db3
• gp3bayespy-0.5.0.tar.gz
  • SHA-256: 95072411316d8d109ba65ddf24ee35c2f672b7276fe3343652aeb2212ebe8719

Governance

The release preserves the gp3bayes governance boundaries: model fitting, diagnostics, predictive scores, sensitivity results, rankings, pupil models, and comparisons do not automatically establish convergence, adequacy, robustness, causality, exclusion decisions, preferred-model status, or psychological/cognitive/emotional interpretation.