docs: correct Dube et al. (2025) JAE citation to issue 7

[igerber]

Corrects the LP-DiD primary-source citation from Journal of Applied Econometrics 40(5):741-758 to the published 40(7):741-758 across all five live sites: docs/references.rst, docs/methodology/REGISTRY.md, docs/methodology/papers/dube-2025-review.md, docs/api/lpdid.rst, and carousel/generate_lpdid_carousel.py. The 3.6.0 release notes retain the original attribution per the repo's historical-CHANGELOG convention; a [Unreleased] → Fixed entry supersedes it.

Methodology references (required if estimator / math changes)

• Method name(s): LP-DiD (citation metadata only — no estimator or math changes)
• Paper / source link(s): Dube, Girardi, Jordà & Taylor (2025), J. Applied Econometrics 40(7):741–758, https://doi.org/10.1002/jae.70000 — issue number verified against the published record (IDEAS/RePEc wly/japmet/v40y2025i7p741-758)
• Any intentional deviations from the source (and why): none

Validation

• Tests added/updated: none needed (docs/comment-string change only; no code paths touched)
• Backtest / simulation / notebook evidence (if applicable): grep -rn '40(5)' docs/ carousel/ diff_diff/ returns zero hits post-change; the only remaining 40(5) is the intentionally retained 3.6.0 CHANGELOG history

Security / privacy

• Confirm no secrets/PII in this PR: confirmed — citation text only

🤖 Generated with Claude Code

https://claude.ai/code/session_012YazJXUeSs5dtLWR5DHpBr

[github-actions]

Overall Assessment

⚠️ Needs changes — no methodology blocker, but the PR introduces a P1 release/version metadata regression.

Executive Summary

• LP-DiD is the only affected method, and the changed estimator content is citation metadata only.
• The 40(7):741-758 citation is consistently reflected in the Methodology Registry and live LP-DiD docs.
• No estimator math, weighting, variance/SE, assumptions, or default behavior changed.
• P1: the PR downgrades package metadata from 3.6.2 to 3.6.1 across Python, Rust, citation metadata, and generated LLM guide files.
• The remaining 40(5) occurrence is in historical CHANGELOG.md release notes and is acceptable per the stated historical-entry convention.

Methodology

Finding: P3 informational — LP-DiD citation metadata corrected
Impact: The Methodology Registry now cites Dube, Girardi, Jordà & Taylor (2025) as Journal of Applied Econometrics 40(7):741-758, matching the updated LP-DiD API/reference docs. No methodology deviation is introduced.
Concrete fix: None required. References: docs/methodology/REGISTRY.md:L1801, docs/api/lpdid.rst:L64, docs/references.rst:L262, docs/methodology/papers/dube-2025-review.md:L4-L16.

Code Quality

Finding: No code-quality issue in estimator logic
Impact: The only Python source change is __version__; no inference, weighting, aggregation, or estimator code path is modified. Known anti-patterns such as inline inference computation and incomplete estimator parameter propagation are not introduced.
Concrete fix: None.

Performance

Finding: No performance issue
Impact: Citation and metadata changes do not affect runtime behavior.
Concrete fix: None.

Maintainability

Finding: P1 — version metadata is downgraded from 3.6.2 to 3.6.1
Impact: A build from this PR would report/package itself as an older already-released version, which can create duplicate release artifacts, package-index upload failures, incorrect diff_diff.__version__, and stale citation metadata. This is outside the stated citation-only scope.
Concrete fix: Restore the intended current version metadata, or split the release-state rollback into a separate explicit release-maintenance PR. Affected locations: pyproject.toml:L7, diff_diff/__init__.py:L304, rust/Cargo.toml:L3, CITATION.cff:L10-L11, diff_diff/guides/llms-full.txt:L5, CHANGELOG.md:L8, CHANGELOG.md:L2093.

Tech Debt

Finding: No untracked deferred-work issue
Impact: No new deferrable implementation limitation is introduced.
Concrete fix: None.

Security

Finding: No security issue
Impact: No secrets, credentials, PII, or unsafe code paths are introduced by the changed files.
Concrete fix: None.

Documentation/Tests

Finding: P3 informational — tests are not necessary for the citation-only docs edits
Impact: The citation correction is covered by direct textual consistency. The only remaining 40(5) occurrence is in historical CHANGELOG.md:L282, which is documented as intentionally retained.
Concrete fix: None for tests. Keep the historical changelog entry if that is the repo convention.

Path to Approval

1. Restore 3.6.2 metadata in pyproject.toml, diff_diff/__init__.py, rust/Cargo.toml, CITATION.cff, and diff_diff/guides/llms-full.txt, unless there is an explicit release-management reason handled in a separate PR.
2. Restore the CHANGELOG.md 3.6.2 heading/link, or split the new citation fix into an [Unreleased] section without deleting/downgrading the existing release metadata.

[igerber]

/ai-review

[github-actions]

🔁 AI review rerun (requested by @igerber)

Head SHA: 9a4ed4a4428466307a2fdfc10f29b2e791d9ac8c

---

Overall Assessment

✅ Looks good — no unmitigated P0 or P1 findings.

Executive Summary

• Affected method: LPDiD; the PR changes citation metadata only.
• No estimator math, weighting, variance/SE, identification assumptions, defaults, or inference paths changed.
• The prior P1 version-metadata regression is resolved: project metadata remains 3.6.2.
• The remaining 40(5) occurrence is only in the historical 3.6.0 changelog entry and is explicitly superseded by the new [Unreleased] note.
• One minor documentation-formatting nit is noted below; it is not blocking.

Methodology

• Finding: LP-DiD citation metadata is consistently updated.
  • Severity: P3 informational
  • Impact: docs/methodology/REGISTRY.md:L1801, docs/methodology/papers/dube-2025-review.md:L4-L16, docs/api/lpdid.rst:L62-L64, and docs/references.rst:L262 now agree on Journal of Applied Econometrics, 40(7):741-758. No methodology deviation is introduced.
  • Concrete fix: None required.

Code Quality

• Finding: No estimator-code quality issue.
  • Severity: P3 informational
  • Impact: The only Python change is a carousel citation string at carousel/generate_lpdid_carousel.py:L816-L820; no inline inference, parameter propagation, control-group logic, or aggregation path changed.
  • Concrete fix: None required.

Performance

• Finding: No performance issue.
  • Severity: P3 informational
  • Impact: Citation text changes do not affect runtime performance.
  • Concrete fix: None required.

Maintainability

• Finding: Prior version-metadata regression is resolved.
  • Severity: P3 informational
  • Impact: Current metadata remains 3.6.2 in pyproject.toml:L7, diff_diff/__init__.py:L304, rust/Cargo.toml:L3, CITATION.cff:L10, diff_diff/guides/llms-full.txt:L5, and CHANGELOG.md:L20.
  • Concrete fix: None required.

Tech Debt

• Finding: No untracked deferred-work issue.
  • Severity: P3 informational
  • Impact: The PR does not add implementation limitations or deferred estimator behavior.
  • Concrete fix: None required.

Security

• Finding: No security issue.
  • Severity: P3 informational
  • Impact: The diff contains citation/documentation text only; no secrets, PII, credentials, or unsafe execution paths are introduced.
  • Concrete fix: None required.

Documentation/Tests

• Finding: Minor Markdown emphasis nesting in the new changelog entry.

  • Severity: P3
  • Impact: CHANGELOG.md:L11-L12 wraps the whole citation in bold and also uses **7** inside it, which may render the issue number inconsistently depending on the Markdown renderer. This is cosmetic only.
  • Concrete fix: Optional: change 40(**7**):741-758** to 40(7):741-758**, or remove the outer bolding and keep only the issue-number emphasis.

• Finding: Tests are not needed for this docs-only correction.

  • Severity: P3 informational
  • Impact: The retained historical citation at CHANGELOG.md:L283-L284 is mitigated by the new superseding note at CHANGELOG.md:L8-L18.
  • Concrete fix: None required.