Title: Blind source separation for multivariate stationary space-time data
Authors: Christoph Muehlmann - Technical University of Vienna (Austria) [presenting]
Sandra De Iaco - University of Salento (Italy)
Klaus Nordhausen - University of Jyvaskyla (Finland)
Abstract: With advances in modern world technology, huge datasets that show dependencies in space as well as in time occur frequently in practice. As an example, several monitoring stations at different geographical locations track hourly concentration measurements of a number of air pollutants for several years. Such a dataset contains thousands of multivariate observations, thus, proper statistical analysis needs to account for dependencies in space and time between and among the different monitored variables. To simplify the consequent multivariate spatio-temporal statistical analysis it might be desirable to find linear transformations of the original data that result in easy interpretative, spatio-temporally uncorrelated processes that are also highly likely to have real physical meaning. Blind source separation (BSS) is a statistical methodology that is concerned with finding so-called latent processes that exactly meet the former requirements. BSS was already successfully used for sole temporal and sole spatial data, but, it was not yet introduced for the spatio-temporal case. BSS is reviewed and a generalization of BSS for second-order stationary multivariate spatio-temporal random fields (stBSS) is proposed. Two novel estimators (stAMUSE and stSOBI) which solve the formulated problem are also provided.