Skip to main content

Reproducibility and provenance

Two executable notebooks​

The study has 2 complementary implementations:

  • notebooks/00_master_chicago_healthmap_pipeline.py is the governed Marimo pipeline.
  • notebooks/RMD_chm_paper_r_Ashley.Rmd is the R statistical analysis and scientific report.

The R report independently reconstructs the primary pooled descriptive measures and HCS hypertension triangulation. It imports specialized bootstrap, HC3, and spatial results from checksum-verified Marimo artifacts to avoid creating a second undocumented implementation.

Integrity controls​

Every required analytic input is bound by SHA-256 digest. The R render stops if an expected file is missing or differs from the completed Marimo manifest. It does not substitute an alternate file.

The review analysis uses:

  • random seed 20260715;
  • 1000 community-area cluster-bootstrap replicates; and
  • 9999 spatial permutations.

Reproduce the public checks​

uv sync --frozen
uv run ruff check .
uv run mypy src
./scripts/test_public_release.sh
uv run marimo check --strict notebooks/00_master_chicago_healthmap_pipeline.py

The public repository includes a deterministic gzip archive of the aggregate geography-condition-year analytic CSV. Materialize and checksum it with:

./scripts/materialize_release_data.sh

The R report additionally requires the 3 checksum-matching Chicago Health Atlas responses listed in sources/public/CHECKSUMS.sha256. Those source responses are not redistributed because their metadata do not state a redistribution license.

Render the R report​

Rscript -e 'rmarkdown::render(
  "notebooks/RMD_chm_paper_r_Ashley.Rmd",
  output_file = "RMD_chm_paper_r_Ashley.html",
  output_dir = "outputs/statistician-review",
  envir = new.env(parent = globalenv()),
  clean = TRUE
)'

Source and artifacts​

The repository contains aggregate artifacts only. Patient-level data, raw clinical source snapshots, credentials, and identifiers are excluded.