Skip to content

README.md

Latest commit

 

History

History
63 lines (50 loc) · 4.91 KB

README.md

File metadata and controls

63 lines (50 loc) · 4.91 KB
Raw

Examples

Worked examples that show what the skills produce in practice.

Contents

File What it shows
aer-exemplars.md Classic AER and adjacent-top-5 papers mapped to each skill, with openICPSR / Dataverse links
modern-aer-exemplars.md 30+ recent (2018-2025) AER and AEJ papers organized by subfield (Labor, Public, Development, Trade, Macro, IO, Health, Environment, Urban, Education, Finance, Political Economy, Social Networks) plus an Identification-Methods table — all with deposit links
intro-example.md A full Keith Head five-paragraph introduction + 97-word abstract, written to AER house style
results-section-example.md Worked body-section narration — sample funnel, finding-first results paragraphs, magnitude conversions, back-of-envelope calculation, mechanisms organized by channel — plus a table-walk-through counterexample
rebuttal-example.md A complete R&R response letter with cover letter, editor response, three referee responses, and a triage summary table
referee-report-example.md A full internal referee simulation — desk screen, three adversarial reports with distinct priors, rubric-scored editor synthesis, and the routed revise list — run on the same fictional paper before submission
skillopt-routing-scenarios.json SkillOpt-style selection/test routing scenarios used by scripts/run_skillopt_gate.py
replication-package-skeleton/ A deposit-ready directory layout with AEA-compliant README, master script, globals file, data/codebook/source-register.md, and docs/exhibit-register.md. Drop-in starting point for your own openICPSR submission
staggered-did-demo/ Runnable Python/R simulation showing why TWFE fails under staggered adoption with heterogeneous dynamic effects
iv-weak-instrument-demo/ Runnable Python simulation showing why conventional 2SLS inference can over-reject with weak instruments and why Anderson-Rubin inference is safer
rdd-polynomial-demo/ Runnable Python simulation showing why high-order global polynomials mislead in RDD and local-linear rdrobust is safer
synthetic-control-demo/ Runnable Python simulation showing why synthetic-control inference comes from the placebo-in-space permutation distribution, not visual pre-period fit
shift-share-demo/ Runnable Python simulation showing why shift-share/Bartik inference belongs at the shock (industry) level, not the region level — region-clustered SEs over-reject
few-clusters-demo/ Runnable Python simulation showing why a cluster-robust t-test over-rejects with few clusters and the wild cluster bootstrap restores nominal size
multiple-testing-demo/ Runnable Python simulation showing why testing many outcomes inflates the family-wise error rate and how Bonferroni/Holm control it without killing power

How to Use

These examples are reference architectures, not paste-able content. The fictional broadband paper used in intro-example.md and rebuttal-example.md is consistent across both files so you can see how the same project moves from initial draft to R&R revision.

To start your own project:

  1. Pick 2-3 papers from aer-exemplars.md that match your design
  2. Read them as architecture templates (reverse-outline in 30 minutes)
  3. Scaffold replication-package-skeleton/ into your project
  4. Use intro-example.md and rebuttal-example.md as reference for the writing skills (aer-introduction, aer-rebuttal)
python3 scripts/scaffold_project.py skeleton /path/to/new-replication-package

Add --dry-run first if you want to inspect the planned copies.

For runnable demos, install the pinned Python stack in ../templates/python/requirements.txt and run the optional smoke gate from the repository root:

make smoke-examples

The smoke gate executes demo assertions when dependencies are present and skips missing optional stacks by default. Use python3 scripts/run_example_smoke.py --strict-deps before release to fail on any missing dependency or failed demo assertion.

Before copying an example's structure into a manuscript, cross-check the design against ../docs/methods-reference.md and run the ../docs/desk-rejection-audit.md. The examples show architecture; the audit and methods reference decide whether that architecture is defensible for your paper.