Metadata-Version: 2.4
Name: data-supply-chain-workshop
Version: 0.1.0
Summary: Building Fairness Into the Data Supply Chain — workshop toolkit (median imputation)
Project-URL: Homepage, https://github.com/WaverlyRose/data-supply-chain
Author: Waverly Rose Brim
License-Expression: Apache-2.0
License-File: LICENSE
License-File: LICENSE-CONTENT.md
Requires-Python: >=3.11
Requires-Dist: matplotlib<4,>=3.7
Requires-Dist: numpy<3,>=1.24
Requires-Dist: pandas<3,>=2.0
Requires-Dist: scikit-learn<2,>=1.3
Requires-Dist: scipy<2,>=1.10
Description-Content-Type: text/markdown

# data-supply-chain-workshop

Workshop toolkit for **"Building Fairness Into the Data Supply Chain"**
(AfroTech 2026). Public NHANES 1999-2004 cohort bundled; no network needed.

```python
from data_supply_chain.bootstrap import cold_start
S = cold_start()          # 14,788 adults, canonical 70/30 split
```

## Relationship to `data-supply-chain`

Same import name, `data_supply_chain`. **Install one or the other, never both** —
they provide the same top-level package and pip will let the second overwrite the
first's files.

This distribution omits the soft-kNN imputation module and uses column-median
imputation instead. Everything else is identical.

## What that changes

Four numbers in the §9 fairness bundle move in the third decimal:

| | full (`data-supply-chain`) | this distribution |
|---|---|---|
| before — worst-group AUROC / gap | 0.788 / 0.128 | **0.787 / 0.130** |
| after — worst-group AUROC / gap | 0.787 / 0.129 | **0.786 / 0.131** |

Every other published figure is identical: reclassification rates, metformin
masking counts (13 / 11 / 2), national projection (427,489), mortality-model
AUROC (0.9013 / 0.9017), the insurance audit (0.776 / 0.911) and the ship-gate
verdict. The bundle's conclusion is unchanged in both — the worst group stays
Uninsured and the insurance gap does not close.

Check which you are running:

```python
from data_supply_chain.preprocess import distribution_variant
distribution_variant()    # "workshop" or "full"
```

## Licence

Apache-2.0 for code, CC BY 4.0 for workshop content.
