Measuring Local Assortativity in the Presence of Missing Values

Abstract

Assortativity or homophily, the tendency of network nodes to connect with similar nodes, often is an useful property to monitor in social networks. For example, when applied to a demographic social network, assortativity on variables such as ethnicity, education level and income is an indication of social segregation. As, for larger networks, the assortativity can vary over the network and between nodes, it is of interest to calculate the assortativity locally. For each person node a local assortativity is calculated while addressing two practical problems. First we apply a normalisation to the local assortativity score to cope with imbalance in group sizes. Secondly, we address missing values in the group identity, which often is a practical problem. We demonstrate the procedures with a real dataset of the Dutch population.