Phase 2 Report: Correcting Observed Wasting Prevalence for Seasonal Variation Using Nonparametric Modelling
This report, part of phase two of the SEASNUT project, proposes a robust, data-driven protocol to correct observed wasting prevalence for seasonal variation at a global scale. Household surveys often record varying wasting prevalence depending on the month they are conducted, particularly influenced by post-harvest periods (lower prevalence) and hunger periods (higher prevalence). Such seasonal fluctuations pose challenges for accurate monitoring and reporting, including Sustainable Development Goal (SDG) targets.
This report answers the question: can we estimate an adjusted wasting prevalence if the survey was conducted in a different month of the year?
This report uses a nonparametric framework to model monthly, yearly, and area-level variability in wasting prevalence, leveraging historical survey data to produce adjusted estimates.
Results of the Report:
Development of a scalable, additive and multiplicative model for seasonal adjustment.
Processing and curation of country-specific data to visualise corrected wasting trends, as opposed to observed wasting trends.
Analysis of grouped countries based on latitude, temperature, GDP, and disaster records.
Exploration of external factors like temperature, rainfall, and disaster events to contextualise wasting patterns.
Creation of reproducible code and tools for stakeholders to further investigate seasonal effects.
Overall, the report delivers significant insights into country-specific and regional trends, equipping stakeholders with evidence-based tools to address data seasonality challenges. While tailored analysis is needed for local contexts, this work lays the foundation for further research into improving accuracy in correcting global wasting prevalence reporting and SDG monitoring.
