Monitoring Progress towards Sustainable Development Goals with Small Area Estimation over Space and Time
Category: International Association of Survey Statisticians (IASS)
Good health and wellbeing is one of 17 Sustainable Development Goals proposed by the UN as a shared vision of peace and prosperity for all. The second goal, good health and wellbeing for all, remains out of reach for many countries, particularly in developing countries. For instance, childhood malnutrition occurs when a child has poor quality and/or inadequate food. Tragically, while it is preventable, it has both short- and long-term detrimental effects. Almost half of childhood deaths are associated with malnutrition, and most of these occur in developing countries in Africa and Asia.
However, proper evaluation and monitoring of countries’ progress towards relies on the availability of good quality data. Unfortunately, many developing countries’ statistical systems struggle to provide accurate data. Notably, countries with poor data tend to be the same low-income countries with high rates of malnourished children. The scarcity of data is also an issue, with many developing countries incurring large gaps between surveys leading to the urgent need for methods to estimate in both the spatial and temporal dimension.
In developed countries such as Australia, the same problems arise in the context of disparities in health outcomes across large spatial scales. While Australia is one the richest countries in the world, and a majority experience good health, Indigenous communities in particular often experience hardship and health challenges. Furthermore, there are localised differences that lead to geographical and spatial variation in health outcomes.
In this session we will address current efforts to monitor health outcomes using a variety of small area estimation techniques. Das will demonstrate the application of the Monte Carlo Markov Chain (MCMC) algorithm to estimate childhood malnutrition rates in spatio-temporal dimensions for Bangladesh. Aheto will demonstrate the integrated nested Laplace approach (INLA) for geo-statistical modelling of childhood mortality in Ghana. Hogg will bring the perspective of an early career researcher to the challenges of including risk factors in models at highly disaggregated levels.
The discussants, Cramb and Baffour will address challenges and future direction for these models. Cramb will demonstrate how the Cancer Atlas of Australia not only estimates smoothed Bayesian models of cancer rates but also visualises them with measures of uncertainty. Baffour will place these and other small area estimation projects in the world context and point to future challenges and opportunities for small area estimation using Big Data.