Analysis of extreme values in financial time series - and approach via the extrememogram

Specialeforsvar: Mathias Brander Hansen

Titel:  Analysis of extreme values in financial time series - an approach via the extremogram

Abstract: In this thesis we focus on the dependence structures in financial time series. The extremogram introduced by Davis and Mikosch [3] was introduced for the purpose Of measuring extremal dependence and will be our main tool. We will calculate the extremogram for several time series models such as AR, GARCH and SV models via simulations and also real-life data. Finally, we also employ different bootstrapping methods for the construction of confidence bands. We find that the cross-extremograms in particular for real-life data implies some strong extremal dependence in several financial time series.

Vejleder:  Thomas Mikosch
Censor:    Yuri Goegebeur, SDU