|
Abstract:
|
In social network analysis, the observed data is usually some social behavior, such as the formation of groups, rather than explicit network structure. Zhao and Weko (2015) proposed a model-based approach called the hub model to infer implicit networks from grouped observations. Hub models assumed independence among groups, which sometimes is not valid for practical consideration. In this article, we generalize the idea of hub models into the case of time varying grouped observations. Similarly to hub models, we assume the group at each time point is gathered by one leader, but allow dependency among groups in the same time segment. We apply a variant of Expectation-Maximization algorithm -- hard EM for identifying group leaders and apply a label switching technique to optimize Bayesian information criterion for identifying segments. The performance of the new model is evaluated under different simulation settings. We apply this model to a data set of Kibale chimpanzee project.
|
