Overall Assessment
⚠️ Needs changes

Executive Summary

• P1: ContinuousDiD docs describe ATT(d)/ACRT(d) and “overall ATT” without stating the Strong Parallel Trends (SPT) requirement; the Methodology Registry distinguishes PT vs SPT for identification of dose-response and global parameters.
• P2: New guidance omits required D=0 (never-treated) group and balanced panel constraints for ContinuousDiD, which are implementation requirements in the Registry and code.
• Docs-only change; no code paths or tests modified. Edge-case checklist is not applicable.

Methodology

• P1: Missing SPT caveat for dose-response outputs in ContinuousDiD docs. Impact: Users may interpret ATT(d)/ACRT(d)/ATT_glob as identified under PT, contradicting the Registry’s PT vs SPT identification boundary. Fix: Add explicit SPT requirement and clarify overall_att equals ATT^{loc} unless SPT in docs/api/continuous_did.rst:9, docs/api/continuous_did.rst:129, docs/choosing_estimator.rst:66, docs/choosing_estimator.rst:188, aligned with docs/methodology/REGISTRY.md:404 and docs/methodology/REGISTRY.md:424.

Code Quality

• No issues found (docs-only change).

Performance

• No issues found (docs-only change).

Maintainability

• No issues found (docs-only change).

Tech Debt

• No issues found (docs-only change).

Security

• No issues found (docs-only change).

Documentation/Tests

• P2: ContinuousDiD guidance does not mention the required D=0 (never-treated) group and balanced panel constraint. Impact: Users may hit runtime ValueErrors or misapply the estimator. Fix: Add a short “Requirements” note in docs/api/continuous_did.rst:14 or docs/choosing_estimator.rst:188 referencing the D=0 requirement and balanced-panel constraint from docs/methodology/REGISTRY.md:440 and docs/methodology/REGISTRY.md:444.
• No tests changed (docs-only change).

/ai-review

🔁 AI review rerun (requested by @igerber)

Head SHA: f729d69ec84c9a95ed58ba7817d548bfd2195d31

Overall assessment: ✅ Looks good

Executive summary

• Methodology statements on PT/SPT identification and target parameters align with the Methodology Registry and ContinuousDiD docstrings.
• New API page and index wiring appear consistent with the exported API and example usage.
• Documentation omits the time-invariant dose requirement enforced by the estimator; add this to the new data requirements.
• Docs-only change; no tests were run or needed.

Methodology

• P2: The new data requirements omit the time-invariant dose assumption that the estimator enforces. Impact: Users with time-varying dose will hit runtime errors or misuse the estimator. Fix: Add a bullet to docs/api/continuous_did.rst:L31-L34 and a note in docs/choosing_estimator.rst:L195-L200 stating dose must be time-invariant across periods (enforced in diff_diff/continuous_did.py:L201-L207).
• P3: PT/SPT identification statements and parameter naming in the new docs match the registry and code docstrings. Impact: None. Fix: None.

Code Quality

• P3: Docs-only PR; no code paths changed. Impact: None. Fix: None.

Performance

• P3: No runtime-impacting changes. Impact: None. Fix: None.

Maintainability

• P3: No maintainability concerns in documentation changes. Impact: None. Fix: None.

Tech Debt

• P3: No new tech debt introduced. Impact: None. Fix: None.

Security

• P3: No secrets or sensitive data added. Impact: None. Fix: None.

Documentation/Tests

• P3: No tests added (docs-only). Impact: None. Fix: None.

/ai-review

🔁 AI review rerun (requested by @igerber)

Head SHA: 37d2f961ae1697b9f3f164881a1b7863d94e97b7

Overall assessment: Needs changes

• P1 doc mismatch: ContinuousDiD analytical SEs are described as “delta method” but registry + implementation use influence functions; fix wording to avoid methodological misstatement.
• ContinuousDiD assumptions and data requirements added here align with the Methodology Registry (SPT/PT distinction, untreated group, balanced panel, time-invariant dose).
• No estimator code changes; risk is documentation accuracy only.

Methodology

• Severity: P1. Impact: Misstates the default analytical SE method for ContinuousDiD, which can mislead users about inference mechanics and comparability across estimators. Fix: Change the ContinuousDiD row to “Analytical (influence-function-based)” or equivalent (docs/choosing_estimator.rst:L278-L280, docs/methodology/REGISTRY.md:L451-L458, diff_diff/continuous_did.py:L854-L868).

Code Quality
No findings.

Performance
No findings.

Maintainability
No findings.

Tech Debt
No findings.

Security
No findings.

Documentation/Tests
No findings beyond the methodology wording issue above.

/ai-review