Box ch. 2 content into numbered divs and callouts

[d-morrison]

Requested by Ezra · project thread

Part of #77

Before: ch. 2 had 5 definition divs and no numbered results, examples or remarks. Its Technical Points 2.1-2.3 and Fine Points 2.1-2.2 were callouts that mixed definitions, results and proofs, so none of them could be cross-referenced.

After: 17 definitions, 13 propositions (each with a proof), 27 examples and 4 remarks, plus 5 warning callouts, 2 tip callouts and 1 note callout.

• Definitions: experiment, ideal randomized experiment, full and mean exchangeability, crossover experiment, marginal and conditional randomization, positivity, MCAR/MAR, stratification, risk period, IP weighted mean and pseudo-population. Each new definition comes before its first use and is followed by an example.
• Results: association is causation under exchangeability, within a stratum too; under exchangeability, association signals an average effect; what each randomized design guarantees; when a crossover experiment reveals an individual effect; MAR is conditional randomization (with a support condition the book's margin note omits, and a counterexample); standardization identifies the counterfactual mean; the properties of the pseudo-population; IP weighting equals standardization and identifies the counterfactual mean.
• Book boxes: Technical Points 2.1-2.3 and Fine Points 2.1-2.2 are split into these divs, and each div title names its source box, as in the ch. 5 and ch. 6 PRs.
• Ids: existing ids (def-exchangeability, def-conditional-exchangeability, def-identification, def-standardization, def-ip-weights) are kept, because ch. 12 and ch. 17 link to them. New ids that could collide take a -ch2 suffix (def-stratification-ch2, def-positivity-ch2, def-pseudo-population-ch2, prp-pseudo-population-ch2). def-ip-weights still duplicates ch. 12's id until Fix duplicate crossref id #def-ip-weights (chapter 12) #114 merges.

How:

• Each boxed statement states its own conditions: dichotomous variables, consistency, exchangeability, positivity, and finite means.
• Two claims are corrected and marked as the notes' own:
  • An effect in some individuals does not by itself create an association.
  • MAR implies conditional randomization only under a support condition.
• Statements and numbers were checked against the book PDF: Tables 1.1, 2.1 and 2.2, and the page cites. Section 2.2's page range is now pp. 17-19.
• Prose is in our own words. An 8-gram overlap check against the book's ch. 2 text now finds only names, dates and the FRCISTG expansion.
• No theorem div sits inside a callout, .notes block or another div.

Checks:

• quarto render chapters/02-randomized-experiments.qmd succeeds for html, docx and revealjs, with no warnings and no unresolved ?@ references. --to latex also succeeds. The PDF could not be compiled locally because TinyTeX is not installed in this container.
• The gha semantic-line-break check passes on the added lines.
• R spelling finds no new words beyond those already flagged on main, which come from math macros and names.
• Four rounds of adversarial review by a separate Claude (Sonnet) reviewer were run, and every finding was addressed.

🤖 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

---

The title change accurately reflects the content (plugging the random variables $A,L$ into $f[a\mid l]$ to get $f[A\mid L]$). This is a minimal, correctly-scoped fix for a real PDF-build break caused by math in a heading, and it doesn't introduce any new cross-reference, macro, or math-correctness issues — the body text and crossrefs are untouched.

I checked whether other remark/div titles in the file still contain inline math (which would risk the same PDF build failure), and found none in this chapter's titles — this was the only offending heading. No other part of the diff changed since the prior round (08dfa0ff..e4e5296a touches only this one line per git diff --stat), and the prior round's full review (cross-reference integrity, div structure, macro usage, math correctness, lab-manual conventions, citations) already covered everything else in the file with no blocking findings.

Verdict

Ready for merge — the only change since the previously reviewed commit (08dfa0ff → e4e5296a) is a one-line title fix (chapters/02-randomized-experiments.qmd:1082) that removes inline math from a rem-ip-weight-notation heading to fix a PDF build break; the new title is accurate and no other content changed. The prior round's verdict (comment, posted 2026-10-09T11:48:28Z, commit 08dfa0ff) found no blocking issues across the full diff, and that assessment still stands.

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "e4e5296a6c225a860810dde1fb29cdbbb8ce047b",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "The only change since the last reviewed commit is a one-line heading edit in chapters/02-randomized-experiments.qmd removing inline math from the rem-ip-weight-notation title to fix a PDF build failure, and the replacement wording remains accurate to the surrounding proof text.",
  "holistic_assessment": "This single-line title fix introduces no cross-reference, macro, or scope regressions, and the previously reviewed structural, mathematical, and lab-manual compliance checks across the rest of the chapter remain valid and unaffected."
}

Reviewed commit: 4089220

[github-actions]

💰 Cost: $0.1611 (review) — run

[github-actions]

PR Preview Action v1.8.1
🚀 View preview at https://Morrison-Lab.github.io/win/pr-preview/pr-123/
Built to branch gh-pages at 2026-10-09 12:17 UTC. Preview will be ready when the GitHub Pages deployment is complete.