befolo/whei-lite-cameroon-dashboard

WHEI-lite Cameroon: composite ADM1 health access proxy index from open HDX data, with robust fallbacks and a static Plotly dashboard for GitHub Pages. # WHEI-lite Cameroon Portfolio project to build a composite **WHEI-lite** proxy index by Cameroon ADM1 regions from open data, then publish a static dashboard in `docs/index.html` for GitHub Pages. ## Project objective Compute a regional composite score (ADM1) using open proxies and publish a static dashboard with: - Choropleth map (WHEI-lite score) - Top/bottom ADM1 bar chart - Regional table (indicators + score) - Method & limitations section - Data sources section ## Open datasets 1. **Cameroon ADM boundaries (HDX / geoBoundaries package)** - Dataset page: data.humdata.org - Download method: CKAN API `package_show` (resource auto-detection) 2. **Cameroon Healthsites (HDX)** - Dataset page: data.humdata.org - Download method: CKAN API `package_show` (resource auto-detection) 3. **Population ADM1 CSV (HDX resource)** - Direct CSV: data.humdata.org ## WHEI-lite indicators (ADM1 proxies) - `I1_facilities_per_100k = facilities_count / population * 100000` - `I2_activities_coverage_rate = share of facilities with non-empty Activities` - `I3_facility_type_diversity = unique Nature of Facility categories / facilities_count` Optional activity signal: - If `Activities` contains SRH/MNCH keywords (`maternity|anc|family planning|hiv|syphilis|hepatitis|malaria`), compute `sgrh_signal_rate` and use it instead of plain I2 for scoring. ## Composite index method 1. Min-max normalize each indicator to [0,1] with zero-division guard. 2. Weighted average score: - `score_whei_lite = 0.5 * I1 + 0.3 * I2_used + 0.2 * I3` 3. Rank regions descending by score. 4. Optional PCA (`scikit-learn`) is attempted; if not feasible, pipeline continues and dashboard reports `PCA not computed`. ## Robust fallback rules - If HDX boundaries fail: fallback to geoBounda …