Correspondence Analysis of Multirelational Multilevel Networks

Abstract

Social actors are often nested within multiple levels that share several members, giving rise to multimodal data. Such data are complex if the actor-nesting is not mutually exclusive. We use affiliation networks to represent teams and individuals, with links representing team membership; social relations between individuals are represented using one-mode networks. We propose an extension of correspondence analysis to multiple levels, incorporating multiple relations and attributes, and demonstrate it with two illustrative examples. We also show how results serve as an exploratory stepping-stone for generating hypotheses to be tested in a more focused manner using confirmatory techniques such as p*/ERGM.