[d-morrison]

Requested by Ezra · project thread

Important

Stacked on #209, which has to land first. Merge with a merge commit, not a squash.

Before: #thm-risk-decomposition and its two lemmas wrote the conditional mean $\E{Y \mid X}$ as $g^*$.

After: they write it as the true mean function $\mu(X)$, matching sds#326.
#thm-best-predictor now says $g^* \eqdef \mu$, so the corollaries and the exercise keep $g^*$ where they state the risk-minimizing property.

How: the latex-macros submodule is bumped to Morrison-Lab/macros c14884b, which adds \truemean and \truemeanf (macros#225); the long names \truemeanfn and \truemeanfnf are deprecated, so this PR uses the short ones.
\bestpredfn becomes \truemean and \bestpredfnf becomes \truemeanf in _thm-risk-decomposition.qmd, _lem-cross-term-zero.qmd and _lem-prediction-error-split.qmd.
The bump from the old pin also renamed \X1n, \Xx1n and \prodi1n to \Xin, \Xxin and \prodin; _def-independence.qmd and _def-pdf-pmf-cdf.qmd use the new names. Those three were the only removed macros that pds uses.
Checked in Chromium on the preview of the earlier commit: 0 MathJax errors on the Expectation and Independence pages.

🤖 Generated with Claude Code

https://claude.ai/code/session_014bhxtsGQ75Wr4yqcz23NqE

[d-morrison]

The build / build check failed on 8b6c992: the PDF of the independence page stopped on Undefined control sequence \X1n. The latex-macros bump in this PR renamed \X1n, \Xx1n and \prodi1n to \Xin, \Xxin and \prodin, and _def-independence.qmd and _def-pdf-pmf-cdf.qmd still used the old names. 66d8f25 renames them. I compared the macro definitions at the old and new pins, and those three are the only removed names that pds uses. The independence PDF now renders locally.

---

Posted by Claude Code (AI agent), not written by a human.
Generated by Claude Code

---

Generated by Claude Code

[github-actions]

Claude finished review — View run

---

The new macros \truemean and \truemeanf are defined in the bumped submodule at latex-macros/macros.qmd:198-199. No instance of the old \truemeanfn or \truemeanfnf remains in the four edited files. That also covers the \bestpredfn \eqdef \truemean alias in _subfiles/_thm-best-predictor.qmd:6, which still expands to μ. I did not render the site.

Review of eaeaf8e ("Use the short true mean macros")

The commit renames \truemeanfn to \truemean and \truemeanfnf to \truemeanf in _lem-cross-term-zero, _lem-prediction-error-split, _thm-best-predictor and _thm-risk-decomposition. It also bumps the latex-macros submodule to c14884b.

• Macros exist: \truemean and \truemeanf resolve to the same μ and μ(·) as the long forms. The long forms are still defined, so any remaining use elsewhere would not break.
• Math unchanged: only the macro names changed. The proof steps, example numbers and \prederr_{\truemean} subscripts read as before.
• Anchors: no #ids or links were touched.

My earlier suggestion about mixed μ and g* notation still applies and is still non-blocking. \bestpredfn and \bestpredfnf remain in _exr-best-predictor, _sol-best-predictor, _cor-best-predictor-argmin, _cor-best-predictor-unique and the rest of _thm-best-predictor. The alias line at _subfiles/_thm-best-predictor.qmd:6 links the two notations only inside that theorem.

The prior verdict, Ready for merge (comment of 2026-10-18T07:43:54Z), still stands.

Verdict

Ready for merge

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "eaeaf8ef3600c541e24e297a08fca63e4c69c136",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "Checked the truemean and truemeanf replacements in four subfiles against the bumped submodule macros at lines 198-199; all are defined and no long-form names remain in the edited files.",
  "holistic_assessment": "The commit is a mechanical rename matching the submodule bump, with no changed anchors, no math changes, and only the earlier optional notation-consistency suggestion remaining."
}

Reviewed commit: eaeaf8e

[github-actions]

💰 Cost: $0.1903 (review) — run

[github-actions]

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

_{2026-10-18 18:27 UTC}