That’s a solid entrant in the “longest seminar title” contest for 2023, but there’s a lot of detail to convey in this project! I am a contributor to this work, behind A/Prof Bernard Baffour (who presented this seminar), Dr Sumon Das (both of the School of Demography, ANU) and Dr Justice Aheto of Accra University, Ghana. The talk was one of the School of Demography seminar series on Tuesday 18 April.
DHS data from 2014 (Ghana, a new survey is in the field currently) and 2017-8 (Bangladesh) were the core data for this project. We developed a geostatistical model that accounted for spatial correlation as well as the effect of covariates from remote-sensed sources such as climate and vegetation data. The outcome (child undernutrition) was modelled by a binomial distribution and resutls presented in terms of exceedance probabilites. This is because of the goals around hunger set by UN in the Sustainable Development Goals – we can model the probability of particular small domains within the countries targeted exceeding the target for percentage of undernutrition.
This is work in progress and the discussion at the end of the talk raised a number of useful direction for thought. It would be great to include bordering countries in the modelling to allow for porous borders, and it would be great to include a temporal aspect as well as spatial. There were also some really good suggestions around using data on maternal health to correlate with child health along with further detail on the nature of health facilities in regions, not just the travel distance to them.
aliceinstatisticsland 