Ego Networks

Egocentric networks are subsets of a larger network which provide a “local view” by assigning one central node as the ego node. These networks are also known as “perceived networks”, since they allow a large network to be analyzed from the “perspective” of some or all of its nodes, comparing the similarities and differences in local structures. Nodes connected to the “ego” are referred to as “alters”.

Ego networks can be constructed based on the following specifications:

• →

  N-step neighborhood: the maximum distance from the “ego” at which a node is considered an “alter” and included in the subset network

• →

  In / Out Neighborhoods: whether incoming or outgoing nodes from the “ego” are included as “alters” in the subset network. Ego networks may also be undirected.

• →

  Strong / Weak Neighborhoods: if a network has weights or lengths assigned to its edges, then neighborhoods can be constructed by following edges which are greater or smaller than a specific weight or length.

Ego networks are often constructed from survey data, which implicitly contains an ego node for the respondent. They can also be extracted from social or infrastructure networks by choosing an ego node and a set of the specifications above. Typical graph metrics that are calculated for ego networks include: network density, longest and average path, centrality, and size. Analyzing ego networks in a whole population provides micro-level insight into its differentiation and cohesion.

Ego networks are well-suited for analyzing systems where actors are constrained or influenced by their local neighbors. When attempting to understand beliefs and preferences in a social network and how they differ from one another, it is the immediate local connections which are the best predictive of an individual. There is research that profitability of peer-to-peer loans can be predicted by the local borrower networks. Ego network analysis provides the appropriate lens and level of data for effective analysis in both these cases.