University of Washington - Department of Statistics
Social interaction data are data that are generated from the interaction or relationship between two or more actors, thus the observational units are pairs, trios, etc. of actors. This type of data are common in all fields of social science (e.g. political science, sociology, anthropology, and economics) for the interaction of actors is a key element in social science theory. In this talk, I focus on data that arise from measurements made on pairs (dyadic data) of actors, where every ordered pair is observed at regular temporal intervals resulting in social network data for each point in time (longitudinal social network data). Typically social network data are used to study a key social phenomenon, such as trade between nations, in relation to a set of predictor variables while accounting for and learning about the interconnectivity of the actors.
Standard static social relations models account for five types of pairwise network dependencies: (1) same sender, (2) same receiver, (3) same transmitter, (4) observational, and (5) reciprocity. The first half of the talk extends these dependencies by incorporating a time dimension. The extended dependence structure is modeled through a random effects approach involving a set of weakly stationary stochastic processes. I will discuss two different parameterizations of the covariance matrices for these processes: one implying a Markov structure and another for a general structure.
In the second half of the talk, I apply the methodology to two real world applications: (1) international trade, and (2) militarized interstate disputes. Afterwards further extensions to the methodology will be discussed.