Correlation Analysis of Stock Indices Using Empirical Mode Decomposition

This thesis studies dependence of ?nancial time series, represented by stock indices of geographically separated economics. Daily prices of selected stock indices from 01.05.1988 to 20.04.2017 are used for the analysis. In the fi?rst section the dataset is described. The second section goes through methodology for the analysis, including empirical mode decomposition (EMD) algorithm, with the help of which the initial time series can be transformed into a few time series, that are mostly independent on each other, for further analysis (for example, correlation analysis). EMD is an interesting method of signal processing, which may help to look at the time series analysis from a different perspective. The thesis should extend the work of Guhathakurta et al. (2008) and extend the time scale and number of indices. Correlation analysis is also performed on the initial time series and the transformed ones. The aim of this thesis is to prove, whether stock indices of geographically separated economics have similarities in their behavior and test whether they are dependent on each other, by using methods from Guhathakurta et al. (2008) extended by correlation analysis.