Add MIT license for code and CC-BY note for data

[alexstorer]

What & why

The repo had no LICENSE file, so under default copyright no one could legally reuse, share, or build on it. This adds one:

• LICENSE — standard MIT License text for the analysis code (scripts/, notebooks/). Verbatim so GitHub's license picker recognizes it.
• README.md — a new ## License section documenting that code is MIT and the processed data (data/processed/) is CC-BY-4.0, derived from the City of San Francisco's public DataSF dataset (vw6y-z8j6), which remains the authoritative owner.

Code and data get different licenses per standard research-data practice. No code or data files change; the analysis pipeline is unaffected.

This closes the one substantive gap found in a github-for-research health check of the repo — everything else (issue-per-experiment workflow, PRs with Closes #N, Claude attribution, raw/processed split, validation assertions, pinned deps) was already in good shape.

🤖 Generated with Claude Code.

[alexstorer]

On the choice of MIT here — quick comparison of the three permissive options, any of which would work for this repo:

• MIT (this PR): shortest, most permissive — "keep the copyright notice," nothing else. The de-facto default for academic/research code and maximally frictionless for reuse. Right when you have no patent surface and just want minimal legalese.
• BSD-3-Clause: essentially MIT plus a non-endorsement clause ("don't use our name to promote derivatives"). Marginally protects the lab's name; common at universities.
• Apache-2.0: permissive like MIT but adds an explicit patent grant + retaliation clause, a NOTICE-file mechanism, and change-notice requirements. Heavier, but the pick if the code embodies anything patentable or you expect large-scale/corporate adoption where the patent grant matters.

For a small SF-311 analysis repo with no patent surface, MIT is the sensible default — it simply signals "reuse freely, keep attribution."

🤖 Generated with Claude Code.

[alexstorer]

Copyright holder — I used "Stanford GSB DARC" as the entity; worth confirming that's the right holder vs. the individual authors or the formal Stanford entity many university repos use ("The Board of Trustees of the Leland Stanford Junior University"). Check your unit's policy on who owns research-code copyright.

Also note a common misconception: MIT does not legally compel citation — it only requires preserving this notice in copies. If you want to be cited in papers (the academic currency), that's a separate lever: add a CITATION.cff and a citation line in the README. License = reuse terms; citation = scholarly norm.

🤖 Generated with Claude Code.

[alexstorer]

Good call splitting code vs data licenses — they're different legal objects and a single MIT/Apache fits data poorly. On the data choice:

• CC-BY-4.0 (this PR): reuse with attribution. Fine, with one subtlety — the underlying DataSF 311 records are public government data you don't own, so you can't license those; what you can license is the selection/arrangement in data/processed/ (your aggregation). CC-BY on the derived tables, framed that way, is correct.
• CC0 (public-domain dedication): often the preferred choice for research data — it removes attribution-stacking friction when others combine many datasets, and many funders/journals now recommend it.

Either is defensible: CC0 for zero reuse friction, CC-BY if you want credit on the derived tables.

🤖 Generated with Claude Code.