A0152
Title: Network extraction and modelling
Authors: Monica Billio - University of Venice (Italy) [presenting]
Abstract: Multidimensional arrays (i.e. tensors) of data are becoming increasingly available and call for suitable econometric tools. Approaches are first revised for extraction of the network also discussing the importance of topology and structure of the data. A new dynamic linear regression model is then proposed for tensor-valued response variables and covariates that encompasses some well-known multivariate models such as SUR, VAR, VECM, panel VAR and matrix regression models as special cases. For dealing with the over-parametrization and over-fitting issues due to the curse of dimensionality, a suitable parametrization is exploited based on the parallel factor (PARAFAC) decomposition, which enables the achievement of both parameter parsimony and incorporates sparsity effects. The contribution is twofold: first, an extension of multivariate econometric models is provided to account for both tensor-variate response and covariates; second, the effectiveness of the proposed methodology is shown in defining an autoregressive process for time-varying real economic networks. Inference is carried out in the Bayesian framework combined with Monte Carlo Markov Chain (MCMC). The efficiency of the MCMC procedure is shown on simulated datasets, with different sizes of the response and independent variables, proving computational efficiency even with high dimensions of the parameter space. Finally, the model for studying the temporal evolution of real economic networks is applied.
