Robust q-negative Multifractal Detrended Cross-Correlation Coefficient
📄 arXiv:2607.06324 · 📥 PDF · 2026-07-07 · cond-mat.stat-mech
Authors: Thiago B. Murari [arXiv · scholar] , José Fernando F. Mendes [arXiv · scholar] , Hernane B. B. Pereira [arXiv · scholar] , Marcelo A. Moret [arXiv · scholar]
🕰 Orloj analysis
Tento článek navrhuje nový koeficient, Signed Multifractal Detrended Cross-Correlation Coefficient ($ρ_{\mathrm{SMFDCCA}}$), který řeší numerické nestability a interpretační problémy stávajícího koeficientu $ρ_q(n)$ pro záporné řády fluktuací. Nový koeficient je striktně ohraničen, numericky stabilní a odhaluje korelační struktury závislé na škále a amplitudě v časových řadách, jako jsou akciové trhy a teplotní záznamy.
💡 Jde o cenné metodologické vylepšení pro analýzu časových řad, které nabízí zvýšenou stabilitu a interpretovatelnost pro studie křížových korelací v komplexních systémech.
✓ modest_claims, falsifiable
📄 Abstract
The multifractal detrended cross-correlation coefficient $ρ_q(n)$ is widely used to investigate scale-dependent interactions, but its application to negative fluctuation orders is affected by numerical instabilities, unbounded values, and interpretational difficulties. We propose a Signed Multifractal Detrended Cross-Correlation Coefficient, $ρ_{\mathrm{SMFDCCA}}(n,q)$, an amplitude-conditioned correlation observable for multifractal detrended analysis, based on locally normalized detrended correlations and regularized fluctuation amplitudes. The proposed coefficient preserves the sign of local interactions, remains strictly bounded within $[-1,1]$ for both positive and negative values of $q$, and eliminates the corrective procedures required by previous approaches. Validation using independent fractional Gaussian noise confirms the absence of spurious cross-correlations and the numerical stability of the method. Applications demonstrate that the proposed observable resolves how cross-correlations evolve jointly with temporal scale and fluctuation amplitude, revealing scale- and amplitude-dependent correlation structures, including stronger synchronization during large fluctuations in stock-market indices and heterogeneous coupling patterns in temperature records.