Longitudinal social network methods for the educational and psychological sciences

Social network analysis is useful for obtaining a better understanding of antecedents and mechanisms of relationship formation and interactions between individuals in educational and psychological contexts. Research utilising descriptive and cross-sectional applications of network analysis is regularly reported, but longitudinal analyses of networks have received less scrutiny. In this methodological article, we compare three commonly applied approaches for analysing longitudinal social network data: Multiple Regression Quadratic Assignment Procedure (MRQAP), Separable Temporal Exponential Random Graph Models (STERGM), and Stochastic Actor Oriented Modelling (SAOM) with research questions about correlations, social structures and mechanisms, respectively. We highlight advantages and disadvantages of the methods and illustrate differences between these methods by analysing longitudinal peer-communication network data of pre-service teachers. The key considerations by the researcher are summarised as ‘FACTS’ (Focus, Assumptions, Conceptualisation, Time points, and Size) as an aid to researchers in selecting the most appropriate method for the analysis of longitudinal social network data.