docs(tutorial-22): align last two survey/weights phrasings with the 3.7.x removal

[igerber]

Summary

• Two markdown cells in tutorial 22 (flagged by the final codex round on refactor(had): remove deprecated survey=/weights= kwargs from the 7 pretest helpers (3.7.x) #659, which merged before the fix could land) still said "permanently deferred under survey/weights" and "QUG under survey / weights is permanently deferred"; both now say "survey designs," matching the post-3.7.x contract where survey_design= is the sole weighting entry. Prose-only — no code, no executed outputs touched; the test_t22_had_survey_design_drift.py guard suite passes unchanged (32), and a repo-wide loose-pattern grep (survey/weights, survey / weights) across all notebooks now returns zero.

Methodology references (required if estimator / math changes)

• Method name(s): N/A — tutorial prose only
• Paper / source link(s): N/A
• Any intentional deviations from the source (and why): None

Validation

• Tests added/updated: None (t22 drift-guard suite passes unchanged).
• Backtest / simulation / notebook evidence (if applicable): N/A

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. Highest unmitigated issue is a P2 documentation anchor regression.

Executive Summary

• No estimator, variance, SE, weighting, or default-behavior code changed.
• Affected methodology surface is documentation for HeterogeneousAdoptionDiD / HAD pretest workflow under survey_design=.
• The tutorial wording now matches REGISTRY.md: QUG is permanently deferred under survey designs, and survey_design= is the sole weighting entry.
• The reverted in-site methodology docs work is tracked in TODO.md, so the version-skew risk is P3 informational under the project rules.
• One documentation defect remains: REGISTRY.md reintroduces a broken DifferenceInDifferences table-of-contents anchor.
• git diff --check passes; I could not run Sphinx because sphinx is not installed in this sandbox.

Methodology

• Severity: P3 informational
  Impact: The notebook prose aligns with the Methodology Registry. docs/tutorials/22_had_survey_design.ipynb:L119-L126, L744-L747, and L824-L827 say QUG is deferred under survey designs, which matches docs/methodology/REGISTRY.md:L3206-L3207 and L3285-L3302. The in-code result text also uses the same “survey designs” language in diff_diff/had_pretests.py:L731 and the workflow suffix at diff_diff/had_pretests.py:L4008.
  Concrete fix: None required.

Code Quality

• Severity: None
  Impact: No Python implementation code changed. No new inline inference computation, partial NaN guards, parameter propagation, or control-group logic changes were introduced.
  Concrete fix: None.

Performance

• Severity: None
  Impact: No runtime paths changed. Removing myst-parser from docs dependencies may reduce docs-install surface, but does not affect estimator performance.
  Concrete fix: None.

Maintainability

• Severity: P3 informational, tracked in TODO.md
  Impact: docs/api/business_report.rst:L48-L68 and docs/api/diagnostic_report.rst:L14-L23 now point to GitHub main for REPORTING.md, so stable ReadTheDocs users may see methodology docs from a different revision. This is explicitly tracked in TODO.md:L60, so it is deferrable under the review rules.
  Concrete fix: No PR-blocking fix required; later complete the TODO by restoring in-site methodology pages or converting those markdown files to .rst.

Tech Debt

• Severity: P3 informational, tracked in TODO.md
  Impact: The PR reverts the prior MyST-based in-site publishing path by removing myst-parser from pyproject.toml, .readthedocs.yaml, .github/workflows/docs-tests.yml, and docs/conf.py. The follow-up is already documented in TODO.md:L60.
  Concrete fix: None required for this PR.

Security

• Severity: None
  Impact: Changes are docs/config/prose only; no credentials, network calls, or secret-handling paths introduced.
  Concrete fix: None.

Documentation/Tests

• Severity: P2
  Impact: docs/methodology/REGISTRY.md:L10 links DifferenceInDifferences to #differenceinifferences, but the heading is ## DifferenceInDifferences at docs/methodology/REGISTRY.md:L42. This breaks navigation in the methodology source material.
  Concrete fix: Change the link back to #differenceindifferences.

• Severity: P3 informational
  Impact: The two deleted LPDiD autosummary source stubs are still referenced by docs/api/index.rst, but LPDiD remains documented directly in docs/api/lpdid.rst:L68-L105. With Sphinx 6 autosummary generation this should be regenerated during docs builds, but I could not verify the Sphinx build locally because sphinx is unavailable.
  Concrete fix: Run make -C docs html SPHINXOPTS="-W" in CI/local docs env; restore the generated stubs if the build does not regenerate them cleanly.

[github-actions]

🔁 AI review rerun (requested by @igerber)

Head SHA: a5c944fc5038426631cdfe9dbeda1277733ec6a0

---

Overall Assessment

✅ Looks good — no unmitigated P0/P1 findings.

Executive Summary

• This PR is prose-only in docs/tutorials/22_had_survey_design.ipynb.
• The affected method surface is HAD / HAD pretest workflow under survey_design=.
• The changed “survey designs” wording matches the Methodology Registry’s post-3.7.x contract that survey_design= is the sole weighting entry.
• QUG under survey designs is explicitly documented as permanently deferred, so this is not a methodology defect.
• No estimator math, variance/SE, defaults, control-group logic, or inference code paths changed.
• git diff --check passed.

Methodology

• Severity: None
  Impact: The notebook wording at docs/tutorials/22_had_survey_design.ipynb:L745-L746 and docs/tutorials/22_had_survey_design.ipynb:L824-L827 aligns with the Registry: survey_design= is the sole HAD weighting entry, QUG directly rejects survey_design=, and the workflow skips QUG while running the survey-aware linearity family. See docs/methodology/REGISTRY.md:L3206-L3207, docs/methodology/REGISTRY.md:L3285-L3299, and runtime strings in diff_diff/had_pretests.py:L731 and diff_diff/had_pretests.py:L4007-L4008.
  Concrete fix: None.

Code Quality

• Severity: None
  Impact: No Python implementation code changed. No new inline inference computation, partial NaN guard, public parameter propagation issue, or control/comparison-group logic was introduced.
  Concrete fix: None.

Performance

• Severity: None
  Impact: Prose-only notebook change; no runtime path changed.
  Concrete fix: None.

Maintainability

• Severity: None
  Impact: The revised terminology reduces API drift by matching the canonical survey_design= wording documented in docs/methodology/REGISTRY.md:L3207.
  Concrete fix: None.

Tech Debt

• Severity: None
  Impact: No new deferred work or TODO dependency introduced. Existing non-diff occurrences of older “survey/weights” wording are outside the changed notebook scope.
  Concrete fix: None.

Security

• Severity: None
  Impact: Documentation/prose-only change; no credentials, secret handling, network calls, or data access paths added.
  Concrete fix: None.

Documentation/Tests

• Severity: None
  Impact: The changed notebook now matches the current code/registry terminology. The prior registry TOC anchor issue is resolved in the checked source: docs/methodology/REGISTRY.md:L10 points to #differenceindifferences, matching the heading at docs/methodology/REGISTRY.md:L42. I did not run the notebook or test suite because this diff changes prose only; git diff --check passed.
  Concrete fix: None.