francescochecchi/acute_malnutrition_predictive_models

R data management and analysis scripts for a study of models to predict the burden of acute malnutrition in South Sudan and Somalia (2014-2018). Somalia datasets are included to enable replication. ## Can we predict the burden of acute malnutrition in crisis-affected countries? Findings from Somalia and South Sudan Checchi et al. (2022) bmcnutr.biomedcentral.com ## Notes on data and R analysis code 23 December 2021 Francesco Checchi, Severine Frison, Abdihamid Warsame Department of Infectious Disease Epidemiology Faculty of Epidemiology and Population Health, Keppel St London School of Hygiene and Tropical Medicine Francesco.checchi@lshtm.ac.uk ### Background on the study This repository contains R scripts and datasets required to replicate a study to develop models to predict the burden of acute malnutrition two crisis-affected countries: Somalia and South Sudan. R scripts are generic, but only Somalia datasets are included. The study was conducted by the London School of Hygiene and Tropical Medicine (www.lshtm.ac.uk) and Unicef, and funded by Unicef and UK Research and Innovation. We combined previously collected datasets of ground anthropometric surveys with a variety of 'predictor' data on factors (food security, conflict intensity, climate, health services, etc.) that are theoretically associated causally with acute malnutrition. We explored both generalised linear models and random forest regressions, and for either of these approaches and a range of anthropometric indicators, quantified various metrics of predictive accuracy. ### Datasets For each of Somalia and South Sudan, the analysis requires the following data files (where 'xxx' = 'som' for Somalia and 'ssd' for South Sudan): * `xxx_analysis_strata_nut.xlsx`, which contains a list of geographic units; * `xxx_predictor_data_nut.xlsx`, which contains all the predictor datasets (oner per worksheet), along with a table of predictors with options for their data management; * `xxx_population_denoms_nut.csv`, which contains population and displacement estimates for each geographic unit and month in the analysis; * `xxx_survey_metadata_nut.xlsx`, which conta …