B1279
Title: Multivariate structural distributional time series
Authors: Simone Maxand - Europa-Universität Viadrina (Germany) [presenting]
Nadja Klein - UA Ruhr and Technische Universitat Dortmund (Germany)
Abstract: A new structural model is proposed for implicit copula time series. Such a generic distributional description of multivariate time series allows studying the interconnection of the involved series in basically all forms while simultaneously capturing the complex and highly non-linear serial dependencies through an implicit copula process. The latter is high-dimensional, but estimation is possible efficiently through Bayesian inference. We illustrate the new method on electricity demand and price data from Germany. We derive electricity price elasticities and short-term probabilistic forecasts for price and demand, both of which are crucial quantities for the efficient operation of energy markets.
