Title: A novel application of spatial statistics in clustering the world's diets
Authors: Thai Le - University of Economics and Business, Vietnam National University Hanoi (Vietnam) [presenting]
Abstract: Notwithstanding the nascent literature on spatial clustering in food economics, previous studies have largely ignored the spatial dimension in clustering dietary patterns. The application of a novel Copula-based K-Medoids Fuzzy Space-Time (COFUST) clustering algorithm is presented for identifying agglomerations of countries that are characterised by similar past trends of food consumption, taking into account their spatial relationship. Specifically, we employ the calorie availability series for 118 countries over the period 1961 to 2013. A key advantage of this approach is the ability to examine the role of the space as a contextual factor for dietary behaviour. The identified clusters not only show similarity in the calorie trajectories but also share homogeneous environment conditions of food consumption. A great novelty is the utilisation of an economic proximity measure instead of traditional metrics for the geographical space. Finally, we introduce the Generalised Fuzzy Morans index that measures the spatio-temporal autocorrelation for spatial units that are collected over time. This index could assist in the selection of the optimal spatial coefficient in the clustering procedure. Results using both simulated and real data show that ignoring the spatial relationship can lead to incorrect interpretation of the clustering results.