Title: An infinite mixture model for clustering of multiplex data
Authors: Silvia DAngelo - University of Rome La Sapienza (Italy) [presenting]
Michael Fop - University College Dublin (Ireland)
Marco Alfo - University La Sapienza, Rome (Italy)
Abstract: Social network analysis is a well-known and growing branch of statistics. Network structures, either single or multivariate, may arise in various contexts and have been investigated in a broad variety of fields. A popular approach to model this type of data is by means of latent variables, which are assumed to influence the observed structure. In particular, latent space models allow us to describe the observed structure by means of an unobserved latent space; units close in the latent space are assumed to be more likely to connect. In many network data, units have the tendency to cluster into communities and this feature has been largely investigated in the context of single networks. A clustering framework for multivariate network data is proposed based on infinite mixtures of Gaussian distributions. We make use of a single latent space and estimate the model parameters within a hierarchical Bayesian framework. A real data application will be presented.