Topological Data Analysis and UNICEF Multiple Indicator Cluster Surveys

Abstract

Multiple Indicator Cluster Surveys (MICS), supported by UNICEF, are one of the most important global household survey programs that provide data on health and education of women and children. We analyze the Serbia 2014–2015 MICS dataset using topological data analysis which treats the data cloud as a topological space and extracts information about its intrinsic geometric properties. In particular, our analysis uses the Mapper algorithm, a dimension-reduction and clustering method which produces a graph from the data cloud. The resulting Mapper graph provides insight into various relationships between household wealth—as expressed by the wealth index, an important indicator extracted from the MICS data—and other parameters such as urban/rural setting, ownership of items, and prioritization of possessions. Among other uses, these findings can serve to inform policy by providing a hierarchy of essential amenities. They can also potentially be used to refine the wealth index or deepen our understanding of what it captures.