Análise multifractal e seções de Lévy de flutuações heterocedásticas

Abstract

An important problem in Physics concerns the study of stochastic processes and fluctuations away from the mean of dynamical variables. In a wide range of systems, some of the observed variables have a macroscopic quality, in the sense that they represent averages or sums over time or space of "microscopic" quantities. When long-range memory or correlation effects do not play a significant role, then the necessary and sufficient conditions for the Central Limit Theorem to hold can become satisfied. Quite often, the second moments of the studied dynamical variable do not diverge, hence in many important instances, the fluctuations of many systems follow Gaussian statistics. On the other hand, complex systems generate some variabilities that often deviate them from Gaussian statistics. Here, we focus on two properties related to Gaussian fluctuations: (i) monofractality and (ii) homoscedasticity. Specifically, we first address the general question about the nature of the relationship between multifractality and heteroscedasticity. We applied multifractal detrended fluctuation analysis to a nonstationary high frequency financial time series obtained from currency markets. As a second test, we applied the technique to the audio time series of Beethoven's fifth symphony. We obtained results suggesting that heteroscedasticity can cause or increase multifractality. We also investigate in greater detail the convergence to the homoskedastic and monofractal Gaussian regime, using the mathematical formalism of Lévy sections, as previously applied to time series. We report several conclusions related to these questions and discuss the generality of these results in the context of the physics of complex systems.