Release v3.1.1

[igerber]

Summary

• Bump version to 3.1.1 across all locations (pyproject.toml, init.py, Cargo.toml, llms-full.txt, CITATION.cff)
• Finalize CHANGELOG entry for v3.1.1

What's in v3.1.1

• Jackknife variance estimation for SyntheticDiD - variance_method='jackknife' implements the delete-one-unit jackknife from Arkhangelsky et al. (2021) Section 5. Supports survey-weighted jackknife with automatic pweight propagation. Validated against R synthdid.
• LinkedIn carousel for dCDH estimator announcement

Methodology references (required if estimator / math changes)

• Method name(s): Version bump only - jackknife methodology reviewed in PR Add jackknife variance estimation to SyntheticDiD #306
• Paper / source link(s): N/A
• Any intentional deviations from the source (and why): None

Validation

• Tests added/updated: No test changes (version metadata only)
• PR Add jackknife variance estimation to SyntheticDiD #306 passed CI

Security / privacy

• Confirm no secrets/PII in this PR: Yes

---

Generated with Claude Code

[github-actions]

Overall Assessment

✅ Looks good

Executive Summary

• This is a metadata-only release PR: version bumps plus a 3.1.1 changelog entry. It does not modify estimator code, inference code, defaults, weighting, or variance computation paths.
• The only method mentioned in the diff is SyntheticDiD, via release notes. That jackknife description matches the existing Methodology Registry and in-code docstring rather than introducing a new undocumented methodology change.
• The edge cases implied by the changelog entry are already covered in surrounding tests: NaN inference for degenerate jackknife cases, analytical p-values, pweight support, R parity, and rejection of full-design survey jackknife.
• Version metadata is consistent across Python packaging, runtime __version__, Rust metadata, citation metadata, and docs. I did not find stale 3.1.0 strings outside historical changelog entries.
• No security, performance, or tech-debt concerns are introduced by this diff.

Methodology

• Severity: P3. Impact: Informational only. SyntheticDiD is the only method referenced in the diff, and only in release notes. The new changelog text is consistent with the documented jackknife method already recorded in the registry and class docstring: fixed-weight leave-one-out jackknife, analytical p-values, and documented survey-weight behavior/restrictions. References: CHANGELOG.md:L10-L14, docs/methodology/REGISTRY.md:L1475-L1531, diff_diff/synthetic_did.py:L50-L64. Concrete fix: None required.

Code Quality

• No findings. The executable-facing changes are limited to straightforward version string updates in metadata/runtime surfaces: pyproject.toml:L5-L8, diff_diff/init.py:L231, rust/Cargo.toml:L1-L4.

Performance

• No findings. No runtime code paths or numerical routines changed.

Maintainability

• Severity: P3. Impact: Informational only. Release metadata is internally consistent across packaging, runtime, citation, and documentation surfaces, which reduces version drift risk during release/publishing. References: pyproject.toml:L5-L8, diff_diff/init.py:L231, rust/Cargo.toml:L1-L4, CITATION.cff:L9-L12, docs/llms-full.txt:L1-L6. Concrete fix: None required.

Tech Debt

• No findings. The PR does not add new deferred work and does not interact with TODO-tracked correctness limitations.

Security

• No findings. The modified files contain no secrets, credentials, or risky executable changes.

Documentation/Tests

• Severity: P3. Impact: Informational only. The new changelog asset reference resolves to an existing file, and the methodology it mentions is already covered by surrounding tests rather than newly introduced here. References: CHANGELOG.md:L13-L14, carousel/diff-diff-dcdh-carousel.pdf, tests/test_methodology_sdid.py:L552-L632, tests/test_methodology_sdid.py:L694-L819, tests/test_survey_phase5.py:L212-L222. Concrete fix: None required.