CMStatistics 2021: Start Registration
View Submission - CMStatistics
B1553
Title: Bayesian mixed-effect models for independent dynamic social network data Authors:  Fabio Vieira - Tilburg University (Netherlands) [presenting]
Joris Mulder - Tilburg University (Netherlands)
Roger Leenders - Tilburg University (Netherlands)
Daniel McFarland - Stanford University (United States)
Abstract: The development of technological devices and communication applications has changed the way humans interact and provide vast amounts of data. As a result, relational event data or timestamped social network data, have been increasingly available over the years. Late developments in statistical modeling of such data focus on methods based on log-linear models. The goal is to model the rates of interactions among actors in a social network via actor covariates and network statistics. The use of survival analysis concepts has allowed the treatment of temporal evolution in social networks. Therefore, more flexible models may be developed with the goal of unveiling the effects driving the network dynamics. We propose a new Bayesian hierarchical modeling approach of independent relational event sequences. This model allows inferences at the actor level, which are useful in understanding which effects guide actors preferences in social interactions. We also present Bayes factor methods for hypothesis testing in this class of models. In addition, a new empirical Bayes factor to test random-effect structures is developed. In this test, we let the prior be determined by the data, alleviating the issue of employing improper priors in Bayes factors and thus preventing the use of ad-hoc choices in absence of prior information, which makes this test quite generally applicable. We use data of classroom interactions among high school students to illustrate the proposed methods.