Disentangling Selection and Causality in Assessing the Effects of Health Inputs on Child Survival: Evidence from East Africa

Abstract

Many demographic data have a hierarchical or clustered structure. For example, the analysis of childhood mortality involves a natural hierarchy where children are grouped within mothers or families, and the latter, in turn, are grouped into communities. Children from the same parents tend to be more alike in their characteristics than children chosen at random from the population at large. To ignore this grouping risks overlooking the importance of group effects, and may render invalid many of the traditional statistical analysis techniques used for studying data relationships.