Release 3.6.0

[igerber]

Summary

• Bump version 3.5.3 → 3.6.0 across diff_diff/__init__.py, pyproject.toml, rust/Cargo.toml, diff_diff/guides/llms-full.txt, and CITATION.cff.
• Promote the CHANGELOG.md [Unreleased] section to [3.6.0] - 2026-06-29 (fresh empty [Unreleased] retained on top) and add the v3.5.3...v3.6.0 comparison link.

This is a release-cut PR: no source/behavior changes of its own — it tags the work already merged to main since v3.5.3. Release highlights:

• New estimator — LPDiD (Local Projections DiD; Dube, Girardi, Jordà & Taylor 2025), top-level export with full API docs.
• PlaceboTests — new optional treatment param on placebo_group_test; fake-group path now filters ever-treated units by default; permutation_test reports the (1+count)/(B+1) randomization-inference p-value.
• ~78–84% peak-memory reduction in the multiplier-bootstrap estimators (Callaway-Sant'Anna, EfficientDiD, HeterogeneousAdoptionDiD) via block-tiled weight generation.
• Methodology validation promoted to Complete: TwoStageDiD, PlaceboTests, Survey Data Support.
• Security: pyo3 + numpy 0.28 → 0.29 (RUSTSEC-2026-0176/0177); CI/deps hardening.
• Numerous perf + rank-deficiency correctness fixes across linalg, imputation, two_stage, conley, spillover, and continuous DiD.

Methodology references (required if estimator / math changes)

• Method name(s): N/A for this PR — version/changelog only. The estimators/fixes it releases were reviewed in their own merged PRs (LPDiD Add LPDiD (Local Projections DiD) estimator (Dube et al. 2025) #575, PlaceboTests feat(diagnostics): PlaceboTests methodology validation (PR-B) — exact RI p-value + R parity #555, etc.), each citing its source in docs/methodology/REGISTRY.md.
• Paper / source link(s): N/A (no methodology code in this diff)
• Any intentional deviations from the source (and why): None

Validation

• Tests added/updated: None (no source changes in this PR; the released work shipped with its own tests).
• Backtest / simulation / notebook evidence (if applicable): N/A
• Note: recommend running the full-matrix wheel dry-run (release-build-check.yml) before merge.

Security / privacy

• Confirm no secrets/PII in this PR: Yes

---

🤖 Generated with Claude Code

[github-actions]

Overall Assessment

✅ Looks good — no unmitigated P0/P1 findings.

Executive Summary

• The diff is release metadata only: changelog header/link, package versions, Rust crate version, CFF release date, and LLM guide version.
• No estimator, inference, weighting, variance/SE, identification, or default-behavior code changed.
• Version surfaces are consistent at 3.6.0; remaining 3.5.3 references are historical changelog entries or the expected comparison base.
• No methodology defects found.
• One minor documentation-maintenance note: the new changelog link reference is inserted out of version order, but it renders correctly.

Methodology

No findings.

• Severity: P3 informational
• Impact: No methodology-affecting files changed. The release note content under CHANGELOG.md:L10 is promoted from [Unreleased] to [3.6.0], but the PR does not alter estimator implementations, assumptions, variance, weighting, or inference paths.
• Concrete fix: None required.

Code Quality

No findings.

• Severity: P3 informational
• Impact: Version constants are updated consistently in diff_diff/__init__.py:L304, pyproject.toml:L7, and rust/Cargo.toml:L3.
• Concrete fix: None required.

Performance

No findings.

• Severity: P3 informational
• Impact: No runtime code paths changed.
• Concrete fix: None required.

Maintainability

Finding: Changelog reference ordering is slightly inconsistent.

• Severity: P3 informational
• Impact: CHANGELOG.md:L1799 adds [3.6.0] after [2.1.0] and before [3.5.3]. Markdown resolution is unaffected, but the reference block is harder to scan.
• Concrete fix: Optionally move [3.6.0] directly above [3.5.3].

Tech Debt

No findings.

• Severity: P3 informational
• Impact: The PR does not add or remove deferred-work entries, and no new untracked correctness limitation is introduced.
• Concrete fix: None required.

Security

No findings.

• Severity: P3 informational
• Impact: No dependency, workflow, credential, or secret-bearing files changed in this diff.
• Concrete fix: None required.

Documentation/Tests

No blocking findings.

• Severity: P3 informational
• Impact: Metadata and docs version surfaces are aligned: CITATION.cff:L10-L11, diff_diff/guides/llms-full.txt:L5, and CHANGELOG.md:L10.
• Concrete fix: None required. Full release-wheel validation is still appropriate before publishing, but this diff does not require new estimator tests.