Dynamic Coupling Index Developed for Heterogeneous Time Series

Summary

The Dynamic Correlation Index (DCI) is a window-based correlation metric that normalizes each signal according to its estimated period before calculating the inner products of the residuals. This approach occupies a middle ground between cheap correlation measures and heavy parametric models, and is used for reconstructing heterogeneous time-series networks.

Why it matters

Many real-world panels, such as EEG, macroeconomic indicators, and social dynamics, contain signals with varying scales, periods, and volatilities. In this regime, Pearson/Spearman methods are insufficient, while VAR/DCC-GARCH models are prone to overfitting or require long data records. The practical appeal of DCI lies in the fact that it adds only one hyperparameter per signal, operates at Pearson-level computational cost, and provides a closed-form null variance for significance testing. This offers a low-friction upgrade for applied researchers. If validation results hold, it could become the default first-pass edge estimator in multi-domain network inference pipelines.

Highlights

• Fills the 'medium-scale data, heterogeneous signal' gap where neither non-parametric correlation nor full parametric models perform well
• Single hyperparameter per signal (estimated period), Pearson-equivalent computational cost, and closed-form null variance for significance testing
• Validated on three external real-world testbeds: synthetic time-varying communities, EEG occipital alpha synchronization, and FRED-MD recession detection

Critical take

The core innovation, period normalization prior to window-based inner products, is a modest extension of Pearson's established estimator rather than a fundamentally new one. Furthermore, since the paper is still in preprint stage (not peer-reviewed), the three validation studies remain single-laboratory, single-run demonstrations lacking independent replication. The inclusion of a biotech company's CEO as a co-author in a statistics/methods-focused paper warrants scrutiny regarding the research context and potential conflicts of interest.