CMStatistics 2017: Start Registration
View Submission - CFE
Title: Copulas for multivariate time series modelling of tail dependence in wind power markets Authors:  Esben Hoeg - Aalborg University (Denmark) [presenting]
Troels Christensen - Aalborg University (Denmark)
Anca Pircalabu - Aalborg University (Denmark)
Abstract: We study the use of copulas in combination with cardinal B-spline wavelets to model multivariate time series. The methods are applied to a bidimensional time series problem within wind power generation.The tail dependences (measures of the strength of dependence in the tails of bivariate distributions) between two important variables at different time scales are estimated, where the series at different time scales are obtained nonparametrically via wavelets. We use copula mixtures for the joint behavior of the individual time series for different time scales. The model is fitted to German data, from where the tail dependences for different time scales are estimated.