In order to investigate the statistical properties of cross-correlations among fluctuations of varying amplitudes in nonstationary data across multiple scales, we extend autoregressive random matrix theory (ARRMT) to multifractal cases and introduce a novel autoregressive random matrix theory approach based on q-Dependent detrended cross-correlation coefficients (q-DCCA-ARRMT). This approach replaces Pearson correlation coefficients with q-DCCA coefficients to construct correlation matrices, thereby integrating multiscale detrending and amplitude-sensitive correlation measures into the ARRMT framework. Numerical experiments demonstrate that the q-DCCA-ARRMT procedure explicitly addresses autocorrelation effects in nonstationary time series, quantifies cross-correlations between small-to-large amplitude fluctuations at distinct temporal scales, and mitigates trend-induced deviation through its detrending technology. The empirical application to 42 constituent stocks of the S&P 500 Index revealed scale-dependent interdependency structures, while the analysis of Beijing’s air quality system identified the dominant pollutants at different time scales. By integrating multifractal analysis with random matrix theory, this framework provides a general tool for characterizing nonlinear, multiscale interactions in complex nonstationary datasets such as those in quantitative finance and environmental science.
扫码关注我们
求助内容:
应助结果提醒方式:
