valentineghanem-bit/gender-equity-literacy-health-ghana

Gender equity, literacy & multi-domain health: spatial ML mediation in Ghana (16 regions / 261 districts) # Gender Equity, Literacy and Multi-Domain Health Outcomes: Spatial Machine-Learning Mediation Analysis in Ghana ## 1. Overview Ecological spatial machine-learning **mediation** study of how **gender equity and female literacy** relate to **multi-domain child and maternal health** (under-five mortality, anaemia, adolescent fertility) across Ghana, and whether **reproductive-health-service use mediates** the link. Inference is at the **16 regions** (2022 DHS); high-resolution spatial structure and mapping use **261 districts** (2021 Census). ## 2. Key findings - Female illiteracy and poverty are **strongly spatially clustered** (Global Moran's I 0.76 and 0.63, both p=0.001), concentrating in the northern savannah; **scale-robust** (16-region 0.80/0.63 — MAUP-checked). - Illiteracy is the **dominant structural correlate** of district poverty (descriptive Random Forest R²=0.88; mean|SHAP| 8.3). - Higher female education/gender equity tracks **lower** under-five mortality (total standardized effect −0.66), child anaemia (−0.96) and adolescent fertility (−0.62) — but **service-mediation is not supported** (all indirect-effect 95% CIs span 0; service coverage already ~89%). - *Associational, ecological, cross-sectional; inference N=16 (exploratory).* ## 3. Study design Ecological cross-sectional, Refined Hybrid (Option C): 16-region DHS inference + 261-district Census spatial layer (3 structural-gap districts merged to parent polygons for rendering only — 261 frame retained). ## 4. Data sources Twelve public datasets: 2022 Ghana DHS subnational indicators (gender, education, literacy, SDG, FP2020, anaemia, child mortality, HIV testing, HIV knowledge), the 2021 Ghana Census 261-district socioeconomic master sheet, the 260-polygon district GeoJSON, and the WHO GHO maternal-reproductive-health series (national context). See `docs/datalog_article16_corrected.md`. ## 5. Repository structure ``` data/{raw,processed} analysis/ src/ outputs/{data,figures,tables} docs …