Box ch. 7 content into numbered divs and callouts

[d-morrison]

Requested by Ezra · project thread

Part of #77

Before: ch. 7 (confounding) had 6 definitions, 1 theorem and 2 examples in numbered divs. Most of the chapter sat in plain prose, speaker notes, or Fine Point and Technical Point callouts. Some terms were used before they were defined, such as "confounding" itself and "no unmeasured confounding".

After: 65 numbered boxes. Each definition comes before its first use and is followed by a concrete example.

• 16 definitions. New ones: confounding (with "common cause"), no unmeasured confounding, M-bias, the open-backdoor-path definition of confounding, signed causal diagrams, surrogate confounder, non-confounder given data on L, g-methods versus conventional methods, negative outcome control, and additive equi-confounding.
• 2 theorems and 5 propositions. Each states its estimand and every condition it uses.
  • Identification under marginal exchangeability.
  • Stratum-specific identification.
  • Standardization (#thm-standardization-ch7, kept; its statement now lists its conditions).
  • Backdoor criterion and conditional exchangeability, with a SWIG proof sketch in 7.5.
  • Adjusting for L alone under the repaired traditional definition (Technical Point 7.2), with a proof that I wrote.
  • Difference-in-differences.
  • The front door formula.
• 29 examples. New ones include conditional exchangeability without the backdoor criterion under an NPSEM-IE, the Technical Point 7.2 conditions checked on Figure 7.4, and a worked aspirin difference-in-differences calculation with hypothetical numbers.
• 13 remarks.
• 7 proofs.
• 19 callouts: 11 notes, 5 warnings, 2 tips and 1 important.

The Fine Point and Technical Point callouts now hold short summaries that point to the boxes after them, as in ch. 6 and ch. 12. No theorem div sits inside a callout or .notes.

How:

• I followed the ch. 8 (Box ch. 8 content into numbered divs and callouts #109), ch. 10 (Box ch. 10 content into numbered divs and callouts #110) and ch. 12 (Box ch. 12 definitions, results and procedures in crossref divs #99) passes and the review findings on Box ch. 1 content into numbered divs and callouts #121, Box ch. 5 content into numbered divs and callouts #118 and Box ch. 6 content into numbered divs and callouts #119.
• I kept the existing ids (#def-backdoor-criterion-ch7, #def-backdoor-path-ch7, #thm-standardization-ch7, #def-sufficient-set, #def-swig, ...) and the section headings, because ch. 8 and ch. 12 link to them. All new ids are unique across the book.
• Claims that exchangeability fails now say "under faithfulness".
• Figures 7.6, 7.11 and 7.12 are not drawn. Their edges are spelled out in the text, checked against the book's figures.

Checks:

• quarto render chapters/07-confounding.qmd succeeds with no warnings to html, docx and revealjs. The page has no unresolved ?@ references.
• A script confirmed that every crossref resolves, no @ reference points forward, no id is duplicated across chapters, and the divs balance. The links into ch. 2, 4 and 6 point to ids or headings that exist.
• The gha check-new-line-breaks and check-informal-definitions checks pass.
• An 8-gram overlap scan against the book chapter finds no added prose shared with the book. The remaining 10-gram matches are citations and the Technical Point title.
• mmHg was added to inst/WORDLIST.
• Five rounds of separate adversarial review (a Claude Sonnet reviewer, since OpenCode is hanging) checked math, conditions, forward references and closeness to the book. All findings were fixed.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Gh7L9vJhwrpvpBhzYMdjPK

---

Generated by Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[github-actions]

Claude finished review — View run

---

Div balance holds (105/105), no duplicate ## box headings anywhere in the chapter, and the old ### section headings remain intact as section titles (now followed by distinctly-titled boxes) with no leftover stray references to the old duplicate phrasing.

Review: PR #124 — follow-up round

What changed since the last review round

Per the computed range b390b5f5..1b64cbe, the only change is commit 76129fe ("ch07: give two boxes titles distinct from their section headings"), a 3-line edit to chapters/07-confounding.qmd touching exactly the two issues flagged in the prior review round (posted 2026-10-09T11:55:46Z, https://github.com/Morrison-Lab/win/actions/runs/37925503433):

1. Duplicate-heading fix (both instances): ## Two Settings in Which the Backdoor Criterion Holds → ## No Confounding, or Confounding That $L$ Removes (line 405), and ## Confounding Is Absolute; Confounder Is Relative → ## Whether a Variable Is a Confounder Depends on the Adjustment Set (line 834). Both now read as distinct titles under their matching ### section headings, matching the pattern used by every other box in this PR. I re-checked div balance (105 open / 105 close) and confirmed no other ## heading duplicates a ### section heading anywhere in the file.
2. #def-confounding precision nit fix: the common-cause definition at line 116 now reads "a directed path to $A$ that does not pass through $Y$ and a directed path to $Y$ that does not pass through $A$" — symmetric and self-contained as requested, rather than relying on the book's implicit $A$-precedes-$Y$ convention.

Both prior findings are resolved exactly as proposed, with no new issues introduced by this small, surgical diff. No other files changed in this range (inst/WORDLIST, macros, crossrefs, etc. are untouched), so none of the previously-verified checks (crossref integrity, cross-chapter anchor slugs, WORDLIST sort order, macro usage) need to be re-run.

Verdict

Ready for merge — both outstanding findings from the prior round are fixed correctly and nothing new was introduced.

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "1b64cbea0b9240577f994a77d47410335e2ad95c",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "Verified chapters/07-confounding.qmd's two renamed box headings no longer duplicate their parent section headings and div balance remains 105 open/105 close.",
  "holistic_assessment": "The 3-line follow-up commit fully addresses both prior-round findings without touching crossrefs, WORDLIST, or macro usage, so regression risk and review scope are both minimal."
}

Reviewed commit: 76129fe

[github-actions]

💰 Cost: $0.2981 (review) — run

[github-actions]

PR Preview Action v1.8.1
🚀 View preview at https://Morrison-Lab.github.io/win/pr-preview/pr-124/
Built to branch gh-pages at 2026-10-09 12:20 UTC. Preview will be ready when the GitHub Pages deployment is complete.