Evaluating probabilistic forecasts of multivariate quantities
The future being uncertain, forecasts ought to be probabilistic in nature, taking the form of probability distribution over future quantities or events.
Frequently, the quantities of interest are multivariate, such as in the prediction of spatio-temporal weather trajectories, multiple asset prices, or population sizes in multiple countries. In view of the ongoing, transdisciplinary transition from point forecasts to probabilistic forecasts, and given that direct analogues of univariate verification techniques tend to lose power in multivariate settings, there is a critical need for the development of theoretically principled and practically useful evaluation techniques for such forecasts, to be addressed in this interdisciplinary workshop.
Prof. Dr. Tilmann Gneiting
Institut für Angewandte Mathematik
Im Neuenheimer Feld 294
Tel: 06221 / 54 4990
Fax: 06221 / 54 5331