Title: Accounting for uncertainty in stochastic actor-oriented models for dynamic network analysis
Authors: Heather Shappell - Johns Hopkins University (United States) [presenting]
Eric Kolaczyk - Boston University (United States)
Abstract: Stochastic Actor-Oriented Models (SAOMs) are designed to capture network dynamics representing a variety of influences on network change in a continuous time Markov chain framework. Developed in the social network setting, these models allow for the testing of hypotheses through the estimation of parameters expressing possible influences on network change. The current framework assumes the observed network edges are free of type I and type II error. However, this is often an unrealistic assumption. We propose a Hidden Markov Model based approach to estimate the error rates and the parameters in the SAOM model. The modeling approach consists of two components: 1) the latent model, which assumes that the unobserved, true networks evolve according to a Markov process as they did in the original SAOM framework; and 2) the measurement model, which describes the conditional distribution of the observed networks given the true networks. An expectation-maximization (EM) algorithm has been developed for estimation, with the incorporation of a particle filtering based sampling scheme due to the enormity of the state space. We present results from a simulation study that demonstrates our method offers great improvement in the accuracy of parameter estimates when compared to the nave approach of just fitting a SAOM. We also apply our method on functional brain networks inferred from electroencephalogram (EEG) data.