Materials
Cheat sheets and curated links for the semester project. Everything here is yours to copy into your own repo — and see templates/ for working versions of the config pattern, cleaning/analysis file skeletons, and the AGENTS.md house-rules file the lecture refers to.
In this folder
| File | When to use it |
|---|---|
| Research decisions log | Every assignment. Log each cleaning/construction choice — and let the agent draft the entries: it made the changes, it writes them up, you review and correct. Documentation gets cheaper with an agent, not more burdensome. |
| Git in VS Code | Finding your way around the Source Control panel — badges, diffs, commits, and the discard button. |
| Git cheat sheet | Keep open until the commands are muscle memory (or skip the memorizing — see below). |
| Good commit messages | When your history starts reading “stuff”, “more stuff”, “final”. |
econ_gitignore_template |
Copy into your repo as .gitignore — keeps data files and log clutter out of Git. |
Three ways to drive Git, pick whichever sticks:
- VS Code’s Source Control panel — the guide above; everything is a button
- GitHub Desktop — a friendly standalone app, nothing to memorize
- Ask the agent — “commit this with a sensible message” works; you’ll still want to read what it committed
Concepts
Short, economics-native reads from Thinking with Agents:
- Terminal Basics — what a shell is and why reproducibility runs through it
- Git & GitHub Essentials — version control from zero
- AI Code Assistance — when AI-generated code is dangerous
- Data Exploration with AI — guardrails for AI + data
LSMS
- LSMS survey catalog — microdata downloads
- LSMS program page — documentation and guidebooks
- LSMS team on GitHub — survey tooling and the harmonized LSMS-ISA agricultural dataset
- Working in R? The
nadaverseandWBqueryRpackages talk to the Microdata Library’s catalog API — good for searching studies and pulling codebooks and file lists programmatically (a nice agent task). The data files themselves you download yourself, through your account (access is requested per survey).
Good research code
- Gentzkow & Shapiro, Code and Data for the Social Sciences: A Practitioner’s Guide — the classic; short and worth reading in full
Tools
- Codex IDE extension docs
- GitHub Student Developer Pack — includes free Copilot
- GitHub Desktop
More
- Thinking with Agents — resources page — a living, annotated list of apps, academic tools, policies, and privacy notes; maintained beyond this course, so link there rather than here for anything current