Institute of Computer Science, Academy of Sciences of the Czech Republic
Pod vodárenskou vezí 2, 182 07 Prague 8, Czech Republic
A method for random resampling of time series from multiscale processes is proposed. Bootstrapped series -- realizations of surrogate data obtained from random cascades on wavelet dyadic trees preserve multifractal properties of input data, namely interactions among scales and nonlinear dependence structures. The proposed approach opens the possibility for rigorous Monte-Carlo testing of nonlinear dependence within, with, between or among time series from multifractal processes.