When Is a Matrix a Geographical Network?

Abstract

Geographical networks are spatial networks in which the nodes have a socially constructed meaning; nodes represent places. All geographical networks can be represented as matrices, but not all matrices in geography represent networks. We argue that a matrix must have at least three properties to represent a geographical network: The rows and columns must represent places that can be associated through the interaction of interest, the entries must represent interactions that have significance beyond dyads, and the values of the entries must be a valid operationalization of the interaction of interest. We illustrate the relevance of the three properties through examples from the city networks literature. These properties serve as guidelines to help geographers determine whether a network analysis of their data is appropriate.