[Zurück]


Diplom- und Master-Arbeiten (eigene und betreute):

G. Kastner:
"Multivariate Generalized Autoregressive Conditional Heteroscedasticity: Theory Multivariate Generalized Autoregressive Conditional Heteroscedasticity: Theory and Application";
Betreuer/in(nen): W. Scherrer; Institut für Wirtschaftsmathematik, 2006.



Kurzfassung englisch:
Time-varying volatility modeling for univariate asset returns is a well investigated topic in time series analysis, including the prominent 2003 Nobel Prize winning ARCH model by Robert F. Engle.

This thesis provides an introduction into general ideas and tools of time series modeling with a special focus on ARCH and generalized ARCH models. Theoretical properties of this model class as well as its fitting to data are discussed. For formulations of generalized ARCH in higher dimensions, several approaches - Constant Conditional Correlation, Dynamic Conditional Correlation, (Diagonal)
Vector GARCH and BEKK models - are presented and analyzed. Because the
curse of dimensionality plays a major role in practical applicability of MGARCH, main attention is placed on models with reasonable numbers of parameters. These are fitted to the Austrian and German stock market indices ATX and DAX making use of S-PLUS with the module finmetrics. Finally, results for empirical prediction performance among the different models are compared by means of both in- and out-sample measurement.

Erstellt aus der Publikationsdatenbank der Technischen Universität Wien.