|
Abstract:
|
Network analyses are popular in representing dependencies and relationships among interacting units. In recent years, multilayer networks have been widely used in understanding connections among a set of nodes with respect to different type of relationships. Our motivation is drawn from social science studies involving dynamic multilayer networks. We are specifically interested in investigating underlying processes in citation network data from JSTOR, where the nodes are authors. Each layer corresponds to a specific type of relationship such as co-authorship, title, etc. We propose a nonparametric dynamic multilayer model which extends Durante and Dunson (2014) to account for multivariate time-series of networks instead of a single dynamic network. Our procedure characterizes edge probabilities by embedding the nodes in a lower-dimensional latent space, and allowing their positions to flexibly change in time and across the different layers. We obtain preliminary results to assess the performance of this model through simulations.
|
