odeymofokeng-wits/wastewater-rk-covid

Regression Kriging for COVID-19 wastewater surveillance in Ekurhuleni, South Africa # Wastewater Surveillance with Regression Kriging **Spatial prediction of COVID-19 cases using wastewater monitoring data in Ekurhuleni, South Africa** This repository contains R code for a geostatistical analysis comparing Regression Kriging (RK) against competing models (OLS Regression, SLMM, Random Forest) for predicting subplace-level COVID-19 incidence from SARS-CoV-2 wastewater surveillance data. ## Overview Municipality-level aggregation of wastewater data dilutes spatial signals. This project implements **Regression Kriging** — a spatial linear mixed model combining trend regression with residual kriging — to improve local predictions and demonstrate superior predictive accuracy over non-spatial alternatives. ## Repository Structure ## Scripts | Script | Description | |--------|-------------| | `script_00_shapefile_exploration.R` | Load and inspect subplace boundary shapefiles | | `script_01_data_loading_cleaning.R` | Import and clean COVID-19 cases, WWTP gene data, vulnerability indices | | `script_02a_spatial_mapping.R` | Create spatial objects and coordinate reference systems | | `script_02b_disaggregation.R` | Assign WWTP gene measurements to subplaces using 3-stage spatial algorithm | | `script_03a_maps.R` | Generate choropleth maps of cases and gene concentrations | | `script_03b_diagnostics.R` | Spatial autocorrelation (Moran's I) and variogram analysis | | `script_03c_temporal.R` | Temporal trends, wave analysis, and cross-correlations | | `script_04_variable_selection.R` | LASSO regularisation and VIF analysis for covariate selection | | `script_05_rk_model.R` | **Core: Week-by-week spatial cross-validation for Regression Kriging** | | `script_06a_slmm_fit.R` | Spatial Linear Mixed Model fitting and diagnostics | | `script_06b_comparison.R` | Benchmark against OLS, SLLM, and Random Forest | | `script_07_evaluation.R` | Model performance metrics and validation | | `script_08_epidemeological.R` | Epidemiological metrics (Rt, CFR, growth rates …