Bump version to 3.5.0

[igerber]

Summary

• Bump version 3.4.2 → 3.5.0 across pyproject.toml, rust/Cargo.toml, diff_diff/__init__.py, diff_diff/guides/llms-full.txt, and CITATION.cff (date-released: 2026-06-01).
• Promote the populated [Unreleased] CHANGELOG block to ## [3.5.0] - 2026-06-01 (fresh empty [Unreleased] retained on top) and add the [3.5.0] comparison link.

This is a release-cut PR — no source/methodology/test logic changes; only version strings and the CHANGELOG section header/link.

Release highlights since v3.4.2 (already merged; documented in the [3.5.0] CHANGELOG block)

• New estimator SyntheticControl — classic SCM (Abadie-Diamond-Hainmueller 2010 / Abadie-Gardeazabal 2003) + in-space placebo permutation inference (Add SyntheticControl estimator (classic SCM, Abadie-Diamond-Hainmueller 2010) — PR-1 core #501, synthetic control: PR-A paper reviews (ADH 2010/2015, Abadie 2021 JEL, CWZ 2021) #497, SyntheticControl: in-space placebo permutation inference + reporting-stack integration #511).
• Methodology-tracker completions: PowerAnalysis (PowerAnalysis methodology review (PR-A): Bloom 1995 + Burlig 2020 source audits #506, PowerAnalysis methodology review (PR-B): Burlig Eq.2 equicorrelated panel variance + tracker Complete #512 — includes a panel-variance behavior change), StaggeredTripleDifference (StaggeredTripleDifference PR-A: Ortiz-Villavicencio & Sant'Anna (2025) paper review #499, StaggeredTripleDifference methodology validation + opt-in Eq-4.14 overall ATT #504), ConleySpatialHAC (ConleySpatialHAC methodology-review-tracker promotion + Bertanha-Imbens 2014 citation correction #496).
• vcov_type Phase 1b completed: EfficientDiD (efficient-did: thread vcov_type as narrow {hc1} contract per Chen-Sant'Anna-Xie 2025 (Phase 1b #4) #495) + TwoStageDiD (TwoStageDiD: thread vcov_type as narrow {hc1} contract (Phase 1b interstitial #5, final) #498); SunAbraham/WooldridgeDiD-OLS vcov_type="conley" (Thread vcov_type="conley" through SunAbraham + WooldridgeDiD (defer StackedDiD) #503).
• Fixes: scale-invariant rank detection + LS solve (Python + Rust); rank-guarded DR/OR influence-function SEs for CallawaySantAnna/TripleDifference/StaggeredTripleDifference (Rank-guarded generalized inverse for DR/OR influence-function SEs (CallawaySantAnna, TripleDifference, StaggeredTripleDifference) #507); Bertanha-Imbens 2014 citation correction.

Methodology references (required if estimator / math changes)

• Method name(s): N/A — no methodology/math code changes in this PR (version bump only). The underlying methodology landed in the feature PRs listed above and is documented in docs/methodology/REGISTRY.md.
• Paper / source link(s): N/A for this PR
• Any intentional deviations from the source (and why): None in this PR

Validation

• Tests added/updated: No test changes in this PR (release cut). Feature PRs efficient-did: thread vcov_type as narrow {hc1} contract per Chen-Sant'Anna-Xie 2025 (Phase 1b #4) #495–WooldridgeDiD: outcome-fit hint for OLS on binary/count outcomes #513 carried their own test coverage, already merged and green on main.
• Backtest / simulation / notebook evidence (if applicable): N/A for this PR
• Note: PowerAnalysis numerical output changes in this release (corrected panel variance / rho-sign) — flagged prominently in the [3.5.0] CHANGELOG entry for downstream users.

Security / privacy

• Confirm no secrets/PII in this PR: Yes

---

Generated with Claude Code

[github-actions]

Overall Assessment

✅ Looks good

No unmitigated P0 or P1 findings. The diff is a release-cut metadata update only in CHANGELOG.md:L8-L10, CHANGELOG.md:L1542-L1542, CITATION.cff:L10-L11, diff_diff/__init__.py:L301-L301, diff_diff/guides/llms-full.txt:L5-L5, pyproject.toml:L5-L8, and rust/Cargo.toml:L1-L4.

Executive Summary

• No estimator, weighting, variance/SE, identification, or default-behavior code changed in this PR, so there is no new methodology-surface change to cross-check against source papers or docs/methodology/REGISTRY.md.
• Version metadata is internally consistent at 3.5.0 across Python packaging, Rust packaging, package __version__, citation metadata, and the LLM guide.
• Release date metadata is consistent at 2026-06-01 in the changed release surfaces (CHANGELOG.md:L10-L10, CITATION.cff:L10-L11).
• The release changelog cut is mechanically correct: [Unreleased] is left empty, 3.5.0 gets its own header, and the new compare link points from v3.4.2...v3.5.0 at CHANGELOG.md:L1542-L1542.
• Existing version propagation remains sound: docs read diff_diff.__version__ at docs/conf.py:L19-L20, and the Rust extension exports CARGO_PKG_VERSION at rust/src/lib.rs:L59-L60.

Methodology

• No findings. Affected methods: none. The PR does not modify estimator implementations, formulas, weighting, variance/SE logic, identification checks, defaults, or docs/methodology/REGISTRY.md.

Code Quality

• No findings. The release bump is consistent across the primary version-bearing files, and a repo-wide search shows no stray 3.4.2 references outside the expected prior-release changelog entries.

Performance

• No findings. No executable code paths or algorithms changed.

Maintainability

• No findings. The current indirection for docs and Rust version exposure continues to reduce drift risk (docs/conf.py:L19-L20, rust/src/lib.rs:L59-L60).

Tech Debt

• No findings. This PR does not introduce new deferred work or require a TODO.md mitigation entry.

Security

• No findings. The diff is metadata/documentation only and contains no secrets, credentials, or new executable surface.

Documentation/Tests

• No findings. The changelog release cut is coherent, and the absence of test changes is appropriate for a metadata-only PR.