foreverherie/Mozambique-malaria-dashboard

For testing # Mozambique Malaria Dashboard. **Geospatial Modelling for Malaria Evidence — Mozambique** An R Shiny teaching dashboard for geospatial analysis of malaria burden across Mozambique's 161 districts, built on routine surveillance data (2017–2024). Modelled after the CEMA Geospatial Platform. Live: --- ## Tech stack ### Runtime | Layer | Package | Role | |---|---|---| | Framework | **shiny** | App server, reactive engine, module system | | UI / theming | **bslib** | Bootstrap 5 theming (`bs_theme`), Google Fonts (Source Sans Pro) | | Mapping | **leaflet** | Interactive choropleth maps of district polygons | | Charts | **highcharter** | Time series, distributions, seasonality, scatter plots | | Tables | **DT** | Interactive data tables with client-side export | | Geospatial | **sf** | GeoJSON I/O, geometry handling, centroids | | Spatial stats | **spdep** | Queen-contiguity neighbours, Moran's I, INLA adjacency graph | | Data wrangling | **dplyr**, **tidyr** | Aggregation, joins, reshaping | | I/O | **readr** | CSV loading | | Formatting | **scales** | Axis/label number formatting | | Utilities | **rlang**, **htmltools** | Tidy eval, HTML generation | ### Frontend assets - **`www/styles.css`** — hand-written CSS (~20 KB); no build step, no bundler - **MathJax 3.2.2** — loaded from CDN for the statistical notation in the lessons - Vanilla JS scrollspy in `app.R` for the workbench anchor navigation ### Modelling The **Areal SAE** tab presents a **BYM2** (Besag–York–Mollié 2) spatial model — a Poisson likelihood with a reparameterised convolution prior splitting spatially structured and unstructured random effects. The model is **fitted offline with R-INLA**, not at runtime. The app loads the precomputed result from `data/bym2_fit.rds` (a list of `fit_summary_fixed`, a 161-row `df` of posterior quantities — relative risks, credible intervals, CV, exceedance probabilities — and `cols`). **INLA is therefore not required to run the app.** It is only needed to re- …