[d-morrison]

Requested by Ezra · project thread

What

Adds \condpopriskf{f}{x}, which prints R(f | x): the expected loss of a model given the input. Ezra chose on 2026-10-28 to distinguish it from the population risk \popriskf{f}, which is its average over the input.

Why

• lds defines only the averaged risk, but its bias–variance theorem works with the pointwise quantity, unnamed.
• Brian Hutchinson's Fall 2025 theory.pdf calls it the "conditional risk" and gives it no symbol.
• Duda and Hart separate the conditional risk from the overall risk, and ESL writes EPE(x₀) beside EPE(f).

Changes

• macros.qmd: \condpopriskf (two required arguments, no optional ones, per the lds parser).
• interpretations.tsv: one row; no duplicate keys.
• notation-choices.qmd: the decision and the other sources' conventions.
• _tests/pdf-include-in-header.qmd: plain, squared, subscripted and tall-argument cases.

Checks

I compiled the test page with pdflatex and read the page image: the exponent rises over the closing parenthesis, including with \frac{a}{b} as the argument. I did not run it in MathJax separately.

Next: an lds PR that adds #def-conditional-risk after this merges and the pin is bumped.

---

Generated by Claude Code

[d-morrison]

/review

---

Generated by Claude Code

[d-morrison]

/review

---

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

---

Generated by Claude Code

[github-actions]

👀 Claude Review Dispatched — run reviewing PR #342 at 7d660aa.

[github-actions]

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

_{2026-11-06 04:29 UTC}

[github-actions]

Claude finished review — View run

---

Verdict

No blocking findings in the diff content reviewed; overall PR status not independently verified by this session. The content review found nothing new: macros.qmd:186, interpretations.tsv:175, notation-choices.qmd:333-337, and _tests/pdf-include-in-header.qmd:42 are byte-identical to the content already reviewed in the prior round at commit 0d2bf104, no duplicate interpretations.tsv keys exist on the current head, and the only new commit (7d660aa) is a non-conflicting merge of main that added nothing to this PR's own diff.

I am not in a position to make a terminal merge-authorization call on this PR. The instrument that can authorize such a call, scripts/check-pr-fully-clean.py, is not available in this session, and this session's gh api access is denied, so I cannot run the paginated commits/<sha>/check-runs or commits/<sha>/status reads either. The only signal I obtained, gh pr checks, showed 2 of 6 checks still pending on the merge commit at the time I checked — a partial count, not a terminal determination.

Whoever advances this PR under the standing merge policy should run python3 scripts/check-pr-fully-clean.py 342 -R Morrison-Lab/macros (via a session with GraphQL access) and confirm exit 0 on head 7d660aaab59ad91cbb9540a0e789507cb73fe575 before merging. This review's scope is limited to the code content, which has no outstanding issues.

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "7d660aaab59ad91cbb9540a0e789507cb73fe575",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "Confirmed the merge commit 7d660aa introduced no changes to macros.qmd, interpretations.tsv, notation-choices.qmd or the pdf-include-in-header test versus the previously reviewed 0d2bf104, and re-ran the duplicate-key check with no collisions found.",
  "holistic_assessment": "Scope stayed limited to the single condpopriskf macro addition with no regression from merged-in upstream PRs, but overall PR status was not independently verified since check-pr-fully-clean.py and paginated check-run endpoints were unavailable in this sandbox."
}

Reviewed commit: 7d660aa

[github-actions]

💰 Cost: $0.3878 (review) — run