Release 3.3.1: HAD survey-design consolidation, dCDH by_path placebos + sup-t bands, Phase 4.5 C

[igerber]

Summary

• Bump version 3.3.0 → 3.3.1 across the 6 version surfaces
• Rename CHANGELOG `[Unreleased]` → `[3.3.1] - 2026-04-25`; add comparison link

Highlights since 3.3.0

• HAD survey-design API consolidation (PR Consolidate HAD survey-design API to single survey_design= kwarg #376) — single `survey_design=` kwarg across all 8 HAD surfaces, matching the rest of the library. `survey=` / `weights=` deprecated with migration guidance; bit-exact regressions guard the legacy paths until removal next minor.
• HAD linearity-family pretests under survey (Phase 4.5 C) — `stute_test`, `yatchew_hr_test`, joint variants, and `did_had_pretest_workflow` now accept survey designs (pweight / PSU / FPC; stratified deferred). `qug_test` permanently rejects survey per Phase 4.5 C0.
• dCDH `by_path` joint sup-t bands (PR Add per-path joint sup-t bands to ChaisemartinDHaultfoeuille.by_path #374) — per-path simultaneous CIs across horizons `1..L_max` within each path; surfaced on `results.path_sup_t_bands` and `to_dataframe(level="by_path")`.
• dCDH `by_path` + placebo (PR dCDH by_path + placebo: per-path backward-horizon placebos (Wave 2 #3) #371) — per-path backward-horizon placebos with R-parity on single-path-cohort panels.
• PanelProfile shape extensions (Wave 2, PR Wave 2: PanelProfile outcome/dose shape extensions + autonomous-guide worked examples #366) — new `outcome_shape` and `treatment_dose` sub-dataclasses; `llms-autonomous.txt` worked-examples section.
• Tutorial 19: dCDH for marketing pulse campaigns (PR Add Tutorial 19: dCDH for Marketing Pulse Campaigns #373/Refocus Tutorial 19 on dCDH alone (drop TWFE comparison) #375).
• SDID placebo R-parity (PR Close SDID placebo R-parity gap: warm-start + R-anchored fixture + test seam #369) — placebo SE now uses R-default warm-start; matches `synthdid:::placebo_se` to within 1e-8 on R's exact permutation sequence.
• GitHub workflow hardening (PR chore(ci): add Dependabot config and workflow permissions #377) — Dependabot config + least-privilege workflow permissions.

Methodology references

• N/A for this PR — version-bump only; no source code, methodology, or Rust changes
• Methodology citations for all underlying work are in `docs/methodology/REGISTRY.md` and the section-level CHANGELOG entries

Validation

• No test changes (release-bump PR)
• All methodology + correctness validation lives in the underlying merged PRs (Wave 2: PanelProfile outcome/dose shape extensions + autonomous-guide worked examples #366, Close SDID placebo R-parity gap: warm-start + R-anchored fixture + test seam #369, HAD Phase 4.5 C: linearity-family pretests under survey #370, dCDH by_path + placebo: per-path backward-horizon placebos (Wave 2 #3) #371, Docs refresh: trim README to a 188-line landing page; redirect contributor conventions #372, Add Tutorial 19: dCDH for Marketing Pulse Campaigns #373, Add per-path joint sup-t bands to ChaisemartinDHaultfoeuille.by_path #374, Refocus Tutorial 19 on dCDH alone (drop TWFE comparison) #375, Consolidate HAD survey-design API to single survey_design= kwarg #376 — see CHANGELOG `[3.3.1]` for the full list)

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. I found one P3 docs/process note.

Executive Summary

• Methods affected: none. This diff is a release-metadata cut, not an estimator/inference change.
• The HAD/dCDH text in CHANGELOG.md is documentation only; this PR does not touch the underlying implementation modules or the methodology registry-backed logic.
• The explicit version surfaces are consistent at 3.3.1: diff_diff/__init__.py:L290, pyproject.toml:L7, rust/Cargo.toml:L3, CITATION.cff:L10-L11, diff_diff/guides/llms-full.txt:L5, and CHANGELOG.md:L8-L14.
• Docs versioning is derived from diff_diff.__version__, so I do not see a missed manual bump in Sphinx config (docs/conf.py:L19-L20).
• No edge-case, NaN/inference, control-group, or parameter-propagation risk is introduced here because none of the touched files participate in runtime estimation paths.
• Runtime import smoke testing was not possible in this environment because numpy is unavailable; assessment is based on static inspection.

Methodology

• Severity: none. Impact: Methods affected: none. The diff only changes release metadata/changelog labeling (CHANGELOG.md:L8-L14, diff_diff/__init__.py:L290, pyproject.toml:L5-L8, rust/Cargo.toml:L1-L4, CITATION.cff:L9-L12, diff_diff/guides/llms-full.txt:L5-L9), so no registry-covered estimator logic, weighting, variance/SE, identification check, or default behavior changed relative to docs/methodology/REGISTRY.md:L1-L25. Concrete fix: none.

Code Quality

• Severity: none. Impact: Version metadata is internally consistent across the changed Python, Rust, packaging, citation, and guide surfaces, which avoids obvious release-drift defects (diff_diff/__init__.py:L290, pyproject.toml:L7, rust/Cargo.toml:L3, CITATION.cff:L10-L11, diff_diff/guides/llms-full.txt:L5). Concrete fix: none.

Performance

• Severity: none. Impact: No executable hot path, numerical routine, or algorithmic branch changed. Concrete fix: none.

Maintainability

• Severity: none. Impact: docs/conf.py derives both release and short version from diff_diff.__version__, so the docs surface stays aligned without another manual edit (docs/conf.py:L19-L20). Concrete fix: none.

Tech Debt

• Severity: none. Impact: This PR does not add new technical debt beyond routine release metadata changes, and nothing here needs TODO tracking. Concrete fix: none.

Security

• Severity: none. Impact: The diff changes only version/release metadata and bundled documentation; no secrets, permissions, or executable security surface changed. Concrete fix: none.

Documentation/Tests

• Severity: P3. Impact: CHANGELOG.md no longer has an [Unreleased] placeholder, while contributor guidance still tells authors to add new bullets under the next unreleased version (CHANGELOG.md:L1-L12, CHANGELOG.md:L1411-L1412, CONTRIBUTING.md:L42). This is a minor workflow/docs inconsistency, not a release blocker. Concrete fix: after cutting 3.3.1, restore an empty ## [Unreleased] section and add an [Unreleased] compare link to HEAD.
• Severity: none. Impact: No new tests were needed for this metadata-only diff. Concrete fix: none.