Diversity of statistical modeling approaches in multilevel social network analysis

Abstract

Diversity of statistical modeling approaches in multilevel social network analysis This article presents some different statistical approaches for multilevel network analysis including the multilevel-p2 models for egocentric network data analysis and the Exponential Random Graph Models (ERGMs) for neo-structural multilevel data. First, we study the stochastic models based on dependence assumptions between dyads, including a hierarchical approach using jointly different networks studies. After presenting the ERGMs, we investigate how multilevel dependence may be introduced in them. We conclude then by the Robins and Wang formalization (2010) which seems to be the most successful attempt to treat the dependencies between the two partially nested networks of different levels.