TY  - THES
T1  - Advances in the Analysis of Energy Commodities and of Multivariate Dependence Structures
A1  - Schlüter,Stephan
Y1  - 2011/03/22
N2  - In the first chapter of the dissertation a new stochastic long-term/short-term model for short-term electricity prices is introduced and applied to four major European indices. Evidence is given that all time series contain certain periodic patterns, and it is shown how to use the wavelet transform for filtering purpose. The wavelet transform is also applied to separate the long-term trend from the short-term oscillation in the seasonal-adjusted log-prices.  Moreover, dynamic volatility is found in all time series, which is incorporated by using a bivariate GARCH model with constant correlation.  The residuals are modeled using the normal-inverse Gaussian distribution.

In the second chapter an overview over different wavelet based time series forecasting methods is given. The methods are tested on four data sets, each with its own characteristics. Eventually, it can be seen that using wavelets does improve the forecasting quality, especially for longer time horizons than one day ahead. However, there is no single superior method; the performance depends on the data set and the forecasting time horizon.

In the third chapter a new formula for extreme Student t quantiles is derived. The derivation is based on the proof for the Gaussian quantile and on the fact that the Student t distribution arises as the limit of a variance-mixture of normals.

In the fourth chapter a theoretical framework and a solved example for valuing a European gas storage facility is presented. For modeling the gas price a mean reverting process with GARCH volatility is chosen. Based on this process dynamic programming methods are applied to derive partial differential equations for valuing the storage facility. As an example a storage site in Epe, Germany, is chosen. In this context the effects of multiple contract types for renting a storage site are investigated and a sensitivity analysis is performed.

In the fifth chapter multivariate copula models are discussed. Using three different four-variate data sets it is analyzed, which model fits best to data sets of dimensions higher than two.

In the last chapter the weak tail dependence coefficient of the elliptical generalized hyperbolic distribution is derived. 

KW  - Zeitreihenanalyse
KW  - Copula
KW  - GARCH
KW  - Prognose
KW  - Wavelets
KW  - Quantile
CY  - Erlangen
PB  - Universitätsbibliothek der Universität Erlangen-Nürnberg
AD  - Universitätsstraße. 4, 91054 Erlangen
L2  - http://www.opus.ub.uni-erlangen.de/opus/volltexte/2011/2435
ER  -