# Upstream reference

This showcase mirrors the April 2026 vintage of the NYC subway
accessibility analysis published as the canonical CASESTUDY of the
[`subway-access`](https://github.com/random-walks/subway-access) v0.5.0
library. The 15 figures and the headline numbers inlined in this
package are pre-rendered snapshots — they are not recomputed by the
notebooks here.

The runnable pipeline that produces them lives upstream:

- **Library**: [`random-walks/subway-access`](https://github.com/random-walks/subway-access)
  — typed Python toolkit for reproducible NYC subway accessibility
  analysis. Version 0.5.0 ships the factor-factory engine-audit
  Appendix D and the jellycell reporting bridge.
- **Example pipeline**: [`examples/accessibility-change-over-time/`](https://github.com/random-walks/subway-access/tree/main/examples/accessibility-change-over-time)
  — `main.py` is the entry point that fetches the MTA + ACS data,
  runs the factor-factory / DiD / spatial pipeline, and emits the
  figures and tables embedded here.
- **Upstream CASESTUDY**: `docs/CASESTUDY.md` in the upstream repo.
- **FOIL request template**: see §3.5 of the upstream CASESTUDY — the
  request chases the 56 missing station upgrade dates from the MTA
  Key Station Program.

## Regenerating against a later vintage

To regenerate against fresh data (e.g., later MTA snapshots or the
2024–2028 ACS vintage):

```bash
git clone https://github.com/random-walks/subway-access
cd subway-access/examples/accessibility-change-over-time
uv sync --extra factor-factory --extra tearsheets
python main.py
```

The output figures, tables, and `CASESTUDY.md` land in `reports/`
under the upstream example directory. To refresh this wrapper:

1. Replace `artifacts/figures/figure-*.png` with the regenerated
   figures (keep the filenames — notebook 02 inlines by exact path).
2. Update the headline-numbers dict at the top of
   `notebooks/01_context_and_pointer.py` to reflect the new vintage.
3. If any numerical result drifts enough to invalidate the cross-walk
   row, update `notebooks/03_cross_walk.py` accordingly.
4. Re-run `pnpm showcase:run showcase-subway-accessibility/notebooks/*.py`
   and `pnpm showcase:render showcase-subway-accessibility` to refresh
   the rendered HTML catalogue.

## FOIL request gap (data provenance caveat)

Of the 157 currently accessible NYC subway stations, 101 have ADA
upgrade years traced to MTA press releases, Governor's announcements,
MTA Capital Program records, Wikipedia station articles, and news
coverage spanning 1987–2026. The remaining 56 — primarily Key Station
Program stations from the 1990s–2010s — use a deterministic hash-based
fallback because per-station completion records are not publicly
documented at time of publication. A FOIL request to the MTA for the
complete Key Station Program completion schedule would close this gap
and enable fully credible DiD identification.

A minimal FOIL request template:

```
To: FOIL Officer, Metropolitan Transportation Authority
Re: FOIL request — Key Station Program completion schedule

Under Article 6 of the Public Officers Law, I request records
identifying the completion date (month and year) of each subway
station ADA upgrade performed under the Key Station Program
(1994–2020). For each station, please provide:

  - Station name and complex ID
  - ADA completion date (month / year)
  - Funding source (Key Station Program, Capital Program cycle, other)
  - Primary accessibility feature completed (elevator, ramp, etc.)

Electronic records (spreadsheet or CSV) preferred. If records are
maintained across multiple offices, please coordinate the response.

Sincerely,
<name> <affiliation>
```

## Version pinning

This wrapper depends on `subway-access >= 0.5, < 0.6` via the
`packages/python-showcase/pyproject.toml` `[subway-access]` optional
extra. The pin intentionally excludes 0.6 because the CASESTUDY
numbers and figure set are expected to refresh at that minor version.
When 0.6 ships, bump the pin and re-run the regeneration recipe above.