In this 18-minute film, John Scott and Mark Tranmer discuss the complementarity of social network analysis and multilevel modelling. Social network analysis offers a way of studying patterns of connectedness, with the ability to identify both people or organizations that are more central to networks, and others that are marginal. Multilevel modelling is a technique for accounting for, and measuring, the extent of variation in outcomes of interest for individual units in population groups. For example: the individual units are pupils, the outcome of interest is examination score, the groups are schools, and the research questions relate to variation in examination scores for pupils within and between schools, given characteristics of the pupils and the schools. The two methods have proved useful to combine in providing answers to questions that have quantitative and qualitative aspects to them, such as the way that networks are embedded in cultures; quantitative measures of network density are usefully complemented by an understanding of the meaning of network ties. In response to questions raised by members of the audience at the event where the conversation took place, the points were made that there is surprising constancy of network structures across time and space, and that the capacity of multilevel models to get beyond aggregates and averages for a given population can be very useful in understanding many areas of social life, including health, and the extent to which individual behaviour might be associated with peer networks in the context of other population groups.

Researchers in conversation: John Scott, University of Essex, and Mark Tranmer, University of Glasgow.