[d-morrison]

Requested by Ezra · project thread

Bumps the latex-macros submodule to macros@80b5089, which merged macros#291.

• The deviation operator is now blackboard bold 𝔻 (\devsym, \deviationf, \devsq, \targeterr).
• The observed deviation (\deviation, \popdev, \cdev) is now blackboard bold lowercase 𝕕 instead of δ.
• Pin change only; no source files change here.
• PDF: the lowercase 𝕕 renders with Quarto's default LuaLaTeX and unicode-math, so this repo needs no extra header. A pdflatex build would need macros-pdf-header.tex from macros.

The pin may also bring in other macros commits since the old pin; the preview and CI show their effect.

---

Generated by Claude Code

[github-actions]

Claude finished review — View run

---

Confirmed: the pinned commit 80b5089 is the tip of macros' main branch, matching .gitmodules' tracked branch. Everything checks out.

---

Review: Bump latex-macros to the split deviation notation (#1418)

This is a submodule-pointer-only PR (latex-macros bumped from 605844f to 80b5089d23a3af203890272a275e8370f2bc6aca, merging Morrison-Lab/macros#291). No .qmd/.R/config files in rme itself change.

Verification performed

• PR status: open, not merged, mergeable, branch is up to date with main. Not a draft; no prior review comments exist on this PR.
• Macro definitions are real, not hallucinated: confirmed by fetching the macros submodule's commit history directly — 80b5089d23a3af203890272a275e8370f2bc6aca is "Split the deviation into a blackboard bold operator and value (add Hoff 1st Course book to Bayesian reading list #291)", currently the tip of macros's main branch (matches .gitmodules's tracked branch), so the pin is sane.
• Notation change matches the PR body exactly: diffed old vs. new macros.qmd. Before: \deviation = \delta, \devn = \deviation (plain δ). After: \devsym = \mathbb{D} (new operator), \devval = \mathbbm{d}, \deviation = \devval, \devn = \deviation (𝕕). \devsq/\targeterr/\deviationf now wrap the operator \devsym instead of the old value macro — consistent with the PR's stated intent and with latex-macros/notation-choices.qmd's own description of the change.
• Impact on rme content: of all the affected macros, only \devn is actually used in this repo, in _subfiles/Linear-models-overview/_def-residual-deviation.qmd and _sec_linreg_diag_residuals.qmd (rendered via chapters/Linear-models-overview.qmd, which is included in the website, book, and handout profiles). \devsq, \ssdev, \targeterr, \deviationf, \deviationv, \smeandev are not used anywhere in rme's own content.
• PDF-rendering claim checked in depth, since it's the one claim with teeth: the PR body says the new lowercase 𝕕 "renders with Quarto's default LuaLaTeX and unicode-math, so this repo needs no extra header," while macros/notation-choices.qmd:379 says the lowercase letter "needs the bbm package... so list [macros-pdf-header.tex] before macros.qmd." These look contradictory at first glance, but they aren't: macros.qmd:1289 adds \providecommand{\mathbbm}[1]{\mathbb{#1}} as a fallback, so when bbm isn't loaded, \mathbbm{d} falls back to plain \mathbb{d}. Quarto's default PDF engine is LuaLaTeX (confirmed against Quarto's docs), which loads unicode-math unconditionally absent mathspec (confirmed against pandoc's fonts.latex template), and unicode-math remaps \mathbb to render lowercase blackboard-bold glyphs from the active math font. Neither _quarto-book.yml nor _quarto-handout.yml sets a custom math font, so Quarto's default math font (which does carry the lowercase double-struck glyphs) applies. Net result: the PDF claim holds for this repo's actual configuration — I could not get a local render to double check end‑to‑end (no quarto binary in this sandbox, and this PR has no preview/build result yet — build was still pending at check time), so this is inference from the macro definitions plus Quarto/pandoc docs rather than an observed render; worth a final visual glance at the book/handout PDF once CI's build job finishes, but I found no reason to expect it to fail.
• LaTeX macro-opportunity check: no new math was added to any rme file in this diff — the only change is the submodule gitlink — so there's nothing in this repo's diff to check for macro-simplification or new-macro candidates. (The actual math changes live in the macros repo's own PR add Hoff 1st Course book to Bayesian reading list #291, out of scope for a per-repo review here.)
• CLAUDE.md / lab-manual compliance: no _subfiles/ edited (so no "clear freezer" label needed, correctly omitted), no workflow/CI files touched, PR is scoped to a single concern (a dependency bump), branch already contains latest main. Nothing here implicates the SERG coding-style/practices manual since no R or application code changed.

Findings

None. I looked hard at the one claim in the PR body that could plausibly be wrong (the PDF no-extra-header claim) and it checks out against the macro definitions, Quarto's documented defaults, and pandoc's template.

Verdict

Ready for merge

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "6635d4482b9f33a7c1e8b8739ca0e66ce4d9510d",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "The only changed path is the latex-macros gitlink in this repo; traced \\devn usage into chapters/Linear-models-overview.qmd and found no breakage from the new blackboard-bold deviation macros.",
  "holistic_assessment": "Verified the PDF rendering claim against Quarto's default lualatex/unicode-math engine and the macros submodule's mathbbm fallback, finding no regression risk for the book or handout build profiles."
}

Reviewed commit: 6635d44

[github-actions]

💰 Cost: $1.1637 (review) — run

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.

_{2026-10-27 21:09 UTC}