Department of Pattern Recognition

(eng): We introduce a new method for detection of long-range cross- correlations and cross-multifractality – multifractal height cross-correlation analysis (MF-HXA). MF-HXA is a multivariate generalization of the height- height correlation analysis. We show that long-range cross-correlations can be caused by a mixture of the following – long-range dependence of separate processes and additional scaling of covariances between the processes. Simi- lar separation applies for cross-multifractality – standard separation between distributional properties and correlations is enriched by division of correlations between auto-correlations and cross-correlations. We further apply the method on returns and volatility of NASDAQ and S&P500 indices as well as of Crude and Heating Oil futures and uncover some interesting results.