Adopt the new potential/observed-outcome macros (\ppi, \opi) in the marginal-risk content

[d-morrison]

Background

The latex-macros submodule was bumped to b1bf86f in #976 (merged), which added:

• \ppi{a} / \potprob{a} → $\pi_a$ (causal marginal risk)
• \opi{a} / \obsprob{a} → $\pi(a)$ (observed marginal risk)
• \potout{a} / \py{a} → $Y_a$ (potential outcome)

These now render book-wide, but nothing uses them yet.

Task

Adopt \ppi/\opi in place of the hand-written \pi_a / \pi(a) in the observed-vs-causal marginal-risk material — primarily the subfiles added in #706:
_sec-causal-marginal-risk.qmd, _sec-g-computation.qmd, _sec-predictive-margins.qmd, _rem-observed-vs-causal-marginal-risk.qmd, and the _thm-*-marginal-risk.qmd / _exm-* files.

Why this is a follow-up (not done in #976)

Those files are added by #706 and are not on main until #706 merges, so the adoption can't be done from a main-based PR. Do this after #706 merges.

Notation decision to make first

\potout/\py render as $Y_a$ (subscript), but the #706 content uses $Y^a$ (superscript, the standard causal-inference convention). Decide whether to:

• (recommended) keep $Y^a$ in the book and adopt only \ppi/\opi (which match $\pi_a$/$\pi(a)$ exactly), leaving the potential-outcome symbol as-is; or
• reconcile — e.g. change the macro to $Y^a$ in d-morrison/macros, then adopt \py too.

Also fold in while doing the macro-adoption pass

Non-blocking macro-style suggestions from #706's review, requiring a new macro in d-morrison/macros (same repo/PR as this adoption work):

• \ratio for OMRR/CMRR. _sec-marginal-rr.qmd:13,29 define \text{OMRR} \eqdef \frac{\pi(1)}{\pi(0)} / \text{CMRR} \eqdef \frac{\pi_1}{\pi_0} with raw \frac, where the existing \ratio/\ratiof macro (already used for \ratio(\odds_1, \odds_2) in _def_OR.qmd) is the idiomatic form for a ratio of quantities. Rewrite as \ratio(\opi{1}, \opi{0}) / \ratio(\ppi{1}, \ppi{0}).
• New \convp macro for convergence in probability. \overset{p}{\to} appears 11 times in Add section to logistic regression on computing marginal risk differences #706's diff / 14 times repo-wide with no covering macro, despite other common relations (e.g. \ind for independence) having one. Add to d-morrison/macros: \def\convp{\overset{p}{\to}}, following the file's existing \def-based relation-macro convention, then replace the raw form across the marginal-risk content (_sec-g-computation.qmd's #thm-g-computation-consistency and its proof, _sec-predictive-margins.qmd's #thm-subgroup-pm, etc.).
• Estimator-indirection consistency (\hb → \eb). _thm-g-computation-consistency.qmd and _sec-predictive-margins.qmd's #thm-subgroup-pm write raw \hat\beta throughout, while _sec-logistic-RR-RD.qmd's conceptual-approach subsection already uses the indirected \eb macro (\def\eb{\est{\beta}}) for the same quantity. Purely a maintainability nit (renders identically) — reconcile to \eb while doing this pass.

Tracked as the deferred adoption half of #975/#976.

[github-actions]

This GitHub issue proposes updating notation in the book's marginal-risk content by adopting new LaTeX macros (\ppi, \opi) for causal and observed marginal risks, introduced in a recent submodule update (#976). These macros standardize rendering of $\pi_a$ and $\pi(a) across the book, replacing manually written equivalents. Implementation is deferred until #706 (which introduces relevant files) is merged, as these files are not yet on main. Additionally, the issue raises a notation decision regarding potential outcomes ($Y_a$ vs $Y^a$), recommending keeping $Y^a$, the standard causal-inference convention, while integrating \ppi and \opi. The task aligns with the deferred adoption tracked under #975/#976.

[d-morrison]

Also fold in a \ratio-macro tidy while adopting the new macros in _sec-marginal-rr.qmd (flagged as an optional, non-blocking style nit by the #706 reviewer):

• _sec-marginal-rr.qmd:13 — \text{OMRR} \eqdef \frac{\pi(1)}{\pi(0)} → \text{OMRR} \eqdef \ratio(\opi{1}, \opi{0})
• _sec-marginal-rr.qmd:29 — \text{CMRR} \eqdef \frac{\pi_1}{\pi_0} → \text{CMRR} \eqdef \ratio(\ppi{1}, \ppi{0})

Per CLAUDE.md's ratio-macro convention (use the generic \ratio/\ratiof when the inputs are the quantities themselves — as with \ratio(\odds_1, \odds_2) in _def_OR.qmd). Doing it here, in the same pass that swaps \pi(a)/\pi_a for \opi/\ppi, avoids a redundant CI/review round on #706 now for a non-blocking style tweak.

---

Generated by Claude Code