blacyn8/kenya-climate-risk

# Spatio-Temporal Bayesian Hierarchical Modeling for Climate-Driven Financial Risk in Kenya **Author:** Blacyn Ochieng **Degree:** BSc in Mathematical Sciences (Statistics Specialization) **Academic Year:** 2026 An advanced, end-to-end statistical computing project that quantifies how localized, lagged geographical climate shocks propagate into commodity price inflation and agricultural economic volatility across Kenyan markets. --- ## 📐 1. Theoretical & Mathematical Framework Unlike traditional Machine Learning models (such as Random Forests or basic OLS regressions) which output a single point estimate and falsely assume data points are independent, this project implements a **Generalized Linear Mixed Model (GLMM)** under a **Log-Normal Likelihood** [🏆]. This mathematically accounts for the positive skewness and heteroskedasticity inherent in financial market prices. ### The Hierarchical Equation: \[Y_{it} \sim \text{Log-Normal}(\mu_{it}, \sigma^2)\] \[\mu_{it} = \beta_0 + \beta_1 X_{i, t-3} + \beta_2 Z_{i, t} + s_i + \gamma_t\] Where: * **\(Y_{it}\)**: The observed commodity market price per kilogram (KES) in county \(i\) during month \(t\). * **\(X_{i, t-3}\)**: The Exogenous Climate Covariate—total monthly rainfall (mm) lagged by exactly 3 months to capture agricultural crop-cycle delay effects [🏆]. * **\(Z_{i, t}\)**: The mean monthly temperature (°C). * **\(s_i\)**: The Spatial Random Intercept representing unique unobserved geographic properties for county \(i\). * **\(\gamma_t\)**: The Temporal Random Intercept accounting for macro-economic currency shocks and baseline inflation shifts over time. --- ## 🛠️ 2. The Full-Stack Technical Pipeline The system bridges foundational school concepts with modern software engineering workflows across a multi-language stack [🏆]: * **Data Ingestion Pipeline**: Messy historical CSV records from the Kenya Meteorological Department (KMD) and the World Food Programme (WFP) are systematically collected and parsed …