Theoretical and empirical models often ignore the interdependencies among the actors and actions that comprise world politics at the same time they seek to understand it. As a result Virtually all extant, empirical work proceeds by assuming the absence of any interdependencies among the actors and interactions. Yet, absent dependencies, these data would not interesting from a world politics perspective. As a way of addressing some of these important dependencies in the network of interactions world politics, we use a latent space approach (Hoff, Raftery, Handcock, 2002).
Specifically, we estimate the number of political interactions among a wide range of Central Asian political actors using a latent space approach applied to a large sample of 36,000 event data. We show that this approach captures important dynamics among the data and extends well beyond propinquity explanations (Tobler's Law) of the sources interaction by constructing latent positions of each actor in an unmeasured social "space." We also show that this approach provides a framework within which missing network data may be usefully and accurately imputed.
