Skip to content

Add LinkedIn carousel for v2.4 (Two-Stage DiD)#162

Merged
igerber merged 1 commit intomainfrom
two-step-did-carousel
Feb 17, 2026
Merged

Add LinkedIn carousel for v2.4 (Two-Stage DiD)#162
igerber merged 1 commit intomainfrom
two-step-did-carousel

Conversation

@igerber
Copy link
Copy Markdown
Owner

@igerber igerber commented Feb 17, 2026

Summary

  • Add 10-slide LinkedIn carousel PDF for the v2.4 release featuring Two-Stage DiD (Gardner 2022)
  • New standalone generator script (carousel/generate_v24_pdf.py) adapted from the v2.3 template
  • LaTeX-rendered equations via matplotlib with Computer Modern font for the math slide
  • Teal accent color (#0891b2) replacing v2.3's amber
  • Slides: hook, recap, TWFE problem diagram, two-stage procedure, math equations, honest SEs comparison, per-observation effects table, code example, full 5x2 toolkit grid, CTA

Methodology references (required if estimator / math changes)

  • Method name(s): N/A - no methodology changes (carousel/marketing assets only)
  • Paper / source link(s): N/A
  • Any intentional deviations from the source (and why): None

Validation

  • Tests added/updated: No test changes (marketing asset, not library code)
  • Backtest / simulation / notebook evidence (if applicable): N/A

Security / privacy

  • Confirm no secrets/PII in this PR: Yes

Generated with Claude Code

10-slide carousel featuring Two-Stage DiD (Gardner 2022) with teal
accent color, LaTeX-rendered equations via matplotlib, and adapted
layouts from the v2.3 carousel template.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@github-actions
Copy link
Copy Markdown

Overall assessment: ⚠️ Needs changes

Executive Summary

  • Cross-checked TwoStageDiD slide content against docs/methodology/REGISTRY.md and diff_diff/two_stage.py; the only mismatch is variance/SE wording that overstates guarantees.
  • Slide 6’s “correct coverage / GMM gets it right” language is stronger than the asymptotic GMM sandwich described in the registry.
  • Minor reliability improvement: ensure temporary equation images are cleaned up even if PDF generation errors.

Methodology
Affected methods: TwoStageDiD (variance/SE messaging), references to TWFE and ImputationDiD.

  • P1 — Impact: Slide 6 labels the GMM sandwich as “correct coverage” and claims “GMM gets it right,” which overstates the asymptotic sandwich variance described in the registry (no finite‑sample adjustment) and could mislead readers into assuming guaranteed coverage. Fix: soften to “asymptotically valid coverage” or “accounts for Stage 1 uncertainty,” and remove the absolute “gets it right” phrasing. (carousel/generate_v24_pdf.py:L599-L684)

Code Quality

  • P3 — Impact: Temporary equation PNGs can be left behind if generation fails before cleanup() runs. Fix: wrap the body of main() in try/finally (or use a context manager) to always call cleanup(). (carousel/generate_v24_pdf.py:L932-L950)

Performance
No findings.

Maintainability
No findings.

Tech Debt
No findings.

Security
No findings.

Documentation/Tests
No findings.

@igerber igerber merged commit a39e0f2 into main Feb 17, 2026
1 check passed
@igerber igerber deleted the two-step-did-carousel branch February 17, 2026 01:59
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant