B1240
Title: Including attributes in dynamic stochastic blockmodels: An application to international trade data
Authors: Silvia Pandolfi - University of Perugia (Italy) [presenting]
Francesco Bartolucci - University of Perugia (Italy)
Maria Francesca Marino - University of Florence (Italy)
Abstract: Dynamic Stochastic Block Models represent a flexible tool for analysis in the presence of longitudinal network data. These models provide a dynamic clustering of network nodes by exploiting a latent variable approach based on a hidden Markov specification. The aim is to simplify the complex data structure typical of networks and extract the relevant information from the data. Motivated by the analysis of network data entailing trade flows among countries recorded in the period 2012-2019, the dyad-independent model is extended by allowing the inclusion of attributes on either the measurement and/or the latent model. Together with information on import-export exchanges, a number of nodal (e.g., the total population and the GDP of the country) and edge (e.g., the existence of a commercial agreement between countries) attributes are indeed available. Including this information in the model allows us a deeper understanding of the determinants of international trade. Due to the intractability of the likelihood function, the estimation of model parameters is performed by relying on a variational approximation, as frequently done in the SBMs framework.
