Getting started for researchers

Pick the path that matches how you work. Researcher-friendly public downloads of calibrated MVP datasets are not shipped as a website product today; do not expect a one-click calibrated CSV from these pages.

Path A — Reading and exploration (no install)

  1. Read what you can do and the diabetes–hospitalization example.
  2. Skim understanding the data for person vs claim grain.
  3. Read interpreting results and limitations before using any numbers in a proposal or talk.

You can learn whether the tool is relevant without writing Python.

Path B — Analyst handoff (tables in R, Stata, Python, …)

  1. Ask a colleague or collaborator who can run the package to export a valid scenario fixture (structural teaching data) via the export and release workflow.
  2. Import the resulting CSV or Parquet files into your analysis environment.
  3. Treat the files as artificial structure for code practice. They are not a calibrated research MVP unless your collaborator also runs the separate Phase 5 worlds workflow and clearly labels that output.

If a separately packaged sample dataset is later approved, this page will list the verified artifact path, supported use, and checksums. That path is not documented as existing functionality today.

Path C — Developer setup

Install locally and use the CLI/API:

Responsible use

Synthetic cohorts—even when adjusted toward selected published statistics—cannot support defensible real-Medicare prevalence, effect, or policy claims. See the FAQ for common boundaries.