Comparative Causal Metrics

An Introduction to Regional Impact Evaluation

Source for the Quarto book by Carlos Mendez. Built with R + Quarto, published to GitHub Pages via Quarto's built-in local-render workflow (quarto publish gh-pages).

• 📖 Read online: https://quarcs-lab.github.io/ccm/
• 📄 PDF: optional download link in the book's navbar (rebuilt on demand)
• 🐙 Source: https://github.com/quarcs-lab/ccm

---

Project status

Stage: seven method chapters drafted with live R against the Proposition 99 dataset (chapter 7 uses the Sakaguchi & Tagawa spatial dataset shipped alongside its replication helpers). Preface and cross-method discussion still to do.

#	Chapter	File	Status
—	Preface	index.qmd	stub
1	Introduction	01-introduction.qmd	first draft (live R; Prop 99 example)
2	Interrupted Time Series	02-interrupted-time-series.qmd	first draft (live R; growth-curve + ARIMA)
3	Regression Discontinuity in Time	03-rd-in-time.qmd	first draft (live R; segmented regression)
4	Basic Differences-in-Differences	04-basic-diff-in-diff.qmd	first draft (live R; CA vs Nevada)
5	Classical Synthetic Control	05-classical-synthetic-control.qmd	first draft (live R; full tidysynth pipeline)
6	Structural Bayesian Time Series	06-structural-bayesian-ts.qmd	first draft (live R; CausalImpact + BSTS)
7	Bayesian Spatial Synthetic Control	07-bayesian-spatial-sc.qmd	first draft (live R; Rcpp + SAR; Sakaguchi & Tagawa replication)
—	References	references.qmd	11 entries (Prop 99 sources ported)

Infrastructure (complete):

• Quarto book project with HTML (auto-published) + LaTeX PDF (on demand)
• Dark/light theme toggle (cosmo / darkly) with custom CSS palette
• renv.lock pinning 146 R packages at R 4.5.2 for reproducibility
• Bibliography wired in (references.bib + apa.csl)
• Cover image and favicon
• Live site on GitHub Pages via quarto publish gh-pages

What the drafted chapters contain: every method chapter (1–6) is built from the same cached Proposition 99 dataset (data/proposition99.rds) with live R code. Chapter 1 frames the potential-outcomes problem, walks the decision tree, and runs a naive pre-post strawman. Chapters 2–6 each take one method family — ITS (growth-curve + ARIMA), RDD on time, single-control DiD, classical Synthetic Control with placebo permutation, and CausalImpact / BSTS — and report the ATT against the same canonical case study. All chunks render live.

Next:

• Preface (index.qmd) — flesh out the front matter.
• Cross-method comparison chapter — bring all six ATT estimates onto one forest plot and discuss the disagreements.
• Per-chapter exercises.

---

Project structure

ccm/
├── _quarto.yml                       # Book config (HTML auto-publishes; PDF on demand)
├── index.qmd                         # Preface (unnumbered)
├── 01-introduction.qmd
├── 02-interrupted-time-series.qmd
├── 03-rd-in-time.qmd
├── 04-basic-diff-in-diff.qmd
├── 05-classical-synthetic-control.qmd
├── 06-structural-bayesian-ts.qmd
├── references.qmd                    # Bibliography target (unnumbered)
├── references.bib                    # BibTeX entries
├── apa.csl                           # Citation style (APA 7)
├── custom.css                        # Theme palette and overrides
├── R/                                # Helpers sourced from chapter setup chunks
│   ├── table_helpers.R               #   gt_pretty() + ms_pretty() for tables
│   └── build_chapter_zips.R          #   pre-render hook → downloads/chapter-NN.zip
├── images/                           # Cover, favicon, figures
├── DESCRIPTION                       # Human-readable dep manifest
├── install_packages.R                # One-time renv bootstrap
├── renv.lock                         # Pinned R deps
└── _book/                            # Render output (gitignored)

Local setup

1. Install Quarto

https://quarto.org/docs/get-started/

2. Install R packages (one-time)

Rscript install_packages.R

This installs renv, then bootstraps the project library and writes renv.lock. Commit renv.lock after it's generated.

On subsequent sessions:

renv::restore()

3. Preview locally

quarto preview

Opens the book in your browser with hot-reload.

4. Render HTML

quarto render --to html

Outputs land in _book/. PDF is on demand only — run quarto render --to pdf when you actually need it. EPUB is no longer produced.

Publishing

The book is published manually via Quarto's built-in gh-pages target — no CI. The standard two-command flow keeps the live site in sync with main:

quarto render --to html
quarto publish gh-pages --no-prompt --no-render

--no-render reuses the freshly built _book/, so HTML chunks only execute once. The publish step then:

1. Switches to a temporary gh-pages worktree.
2. Copies _book/ into it.
3. Force-pushes the rendered output to the gh-pages branch on origin.
4. Returns to your working branch.

GitHub Pages serves the gh-pages branch at the root path. The first publish requires an empty gh-pages branch to exist on the remote and Pages to be configured (Source: gh-pages /) — both already done in this repo. Subsequent publishes are one command.

Tip: if git push fails inside quarto publish with HTTP 400, bump git's HTTP post buffer once:

git config http.postBuffer 524288000

Writing chapters

Each chapter is a .qmd file with R code chunks:

---
title: "Chapter title"
---

## Section

Some prose. Cite like this: [@abadie2003economic].

```{r}
library(tidyverse)
# code
```

Add bibliography entries to references.bib. They appear on the References page automatically.

The preface (index.qmd) and references (references.qmd) use a top-level # Heading {.unnumbered} instead of a YAML title: field, so they appear in the sidebar without a chapter number.

License

MIT — see LICENSE.