Repository navigation
Open
1 / 71 of 7 issues completedLabels
Description
Workstream 3 of the development plan.
Where the content stands
intro-causal.qmddefines actions, potential and counterfactual outcomes, individual effects, consistency and (conditional) exchangeability through thedefinitions/subfiles, and stops there.collider-bias.qmduses an ASCII arrow diagram and cites one web page for nearly every claim.bias-analyses.qmdis a single block quotation.summary.qmdis the Quarto template placeholder and is rendered.- There are no chapters yet on average causal effects, identification, causal diagrams, confounding or selection bias, which the later chapters assume.
- There are no exercises and no drawn diagrams.
What to do
- Agree the outline first (see the scope decision in the parent issue).
A reasonable order: causal questions and effects, identification assumptions, causal diagrams, confounding, selection and collider bias, measurement bias, quantitative bias analysis, then the design-based methods. - Remove the placeholder
summary.qmdand the unusedintro.qmd. - Finish the introduction: average causal effects, effect measures, and the identification conditions with the existing exchangeability results.
Sub-issues: the Neyman-Rubin model (#12), causation without manipulation (#11), "marginally exchangeable" as a synonym (#35), "liberating set" (#16), the Bradford Hill criteria (#21). - Write a causal diagrams chapter with drawn DAGs (
ggdagordagitty), including time in DAGs (#5).
Redraw the collider chapter's diagrams the same way and cite the primary papers beside the Catalogue of Bias page. - Turn the bias analysis quotation into a chapter: what a sensitivity analysis can and cannot show, with a worked example.
- Write the competing-events summary (#29) as its own page.
- Give each definition and result an example and each chapter exercises, as the lab's authoring rules ask.
- Review and land the optional philosophy chapter in #53.
General causal-inference explanations now sitting in project repositories should move here as they are found, rather than being rewritten.
Posted by Claude Code (AI agent) --- not written by a human.
Activity
dem-extra1 commentedon Nov 11, 2026
A notation question for this workstream, found while moving the notes to the shared macros in #60.
The notes write a potential outcome as Y(a):
definitions/_def-potential-outcome.qmdsays "We will use notation Y(a)", and the exchangeability definitions and theorems writeY(x)literally.The shared macros define
\potout{a}, which prints a subscript, Y_a (macros#71).So the notes and the lab's shared notation disagree, and the lab rule is to use the shared macro.
The options:
\potout, so they print Y_a, and reword the definition's notation sentence to match.\potoutinMorrison-Lab/macrosto print Y(a), which changes every site that uses it.I recommend option 1: the macro is the lab's recorded choice,
winfollows a book that uses the superscript form and can map to it the same way, and one edit to the macro later would then change all the notes together.This needs Ezra's choice before anyone edits the chapters.
Until then #60 keeps the literal
Y(a), so it changes no rendered math.Posted by Claude Code (AI agent) --- not written by a human.
Generated by Claude Code