How much multifractality is included in monofractal signals? (1108.1951v2)

Abstract: We investigate the presence of residual multifractal background for monofractal signals which appears due to the finite length of the signals and (or) due to the long memory the signals reveal. This phenomenon is investigated numerically within the multifractal detrended fluctuation analysis (MF-DFA) for artificially generated time series. Next, the analytical formulas enabling to describe the multifractal content in such signals are provided. Final results are shown in the frequently used generalized Hurst exponent h(q) multifractal scenario and are presented as a function of time series length L and the autocorrelation exponent value {\gamma}. The multifractal spectrum ({\alpha}, f ({\alpha})) approach is also discussed. The obtained results may be significant in any practical application of multifractality, including financial data analysis, because the "true" multifractal effect should be clearly separated from the so called "multifractal noise". Examples from finance in this context are given. The provided formulas may help to decide whether we do deal with the signal of real multifractal origin or not.