nhisml 0.4.0

Survey-aware ML starter kit for NHIS Adults (2023/2024): fetch, build core data, train baselines, export reproducible outputs.

nhisml

Survey-aware machine learning toolkit for NHIS Adults data.

Features

- Task-aware training (SRH, smoking, etc.)
- Survey-weighted metrics
- Cross-year evaluation (2023 → 2024)
- Subgroup fairness analysis
- Publication-ready outputs

Installation

pip install nhisml

Quick start

#Download and cache raw NHIS Adults public-use files.
nhisml fetch --year 2023 --year 2024
#Builds a clean, analysis-ready core parquet with harmonized variable names.
nhisml build-core --year 2023

Self-rated health

#Train
nhisml train --in data/core_2023.parquet --task srh_binary
#Evaluate
nhisml evaluate --task srh_binary --latest --year 2024
# Subgroup analysis
nhisml subgroup --task srh_binary --latest --year 2024 --by sex age education

Current smoking

#Train
nhisml train --in data/core_2023.parquet --task smoking_current
#Evaluate
nhisml evaluate --task smoking_current --latest --year 2024
# Subgroup analysis
nhisml subgroup --task smoking_current --latest --year 2024 --by sex age education

FIGURES & TABLES

python scripts/make_paper_outputs.py \
  --tasks srh_binary smoking_current \
  --run-for-task srh_binary=runs/<srh_run_dir> \
  --run-for-task smoking_current=runs/<smoking_run_dir>

Others

nhisml list-tasks # list all available predicton tasks
nhisml describe-task # Describe a specific task.
nhisml list-featuresets # List available feature sets.
nhisml describe-featureset  # Describe feature set contents.

LICENSE

This project is licensed under the MIT License.
See the LICENSE file for details.
...