|
Abstract:
|
While observation driven time series models for binary data have been well studied, this class of models has been slower to develop within the context of network analysis. A major complaint with this class of models for networks is the assumption that the dyads are conditionally independent given the edge histories. This assumption is quite strong and is generally difficult to justify. One would typically expect not only the existence of temporal dependencies through which the network at varying time points are dependent, but also simultaneous dependencies which help determine how the dyads of the network co-evolve. We propose a general observation driven model for dynamic networks which overcomes this problem by modeling both the mean and the covariance structures as functions of the edge histories using a flexible autoregressive approach. We propose a visualization method which provides evidence concerning the existence of simultaneous dependence. We describe a simulation study to determine the method's performance in the presence and absence of simultaneous dependence, and we analyze a friendship network, a proximity network from conference attendees, and world trade.
|
