Finite sample properties of power-law cross-correlations estimators

TL;DR

This paper evaluates the finite sample performance of three estimators for power-law cross-correlations, highlighting their strengths and limitations depending on the time series dynamics.

Contribution

The study provides a comprehensive Monte Carlo simulation analysis of DCCA, HXA, and DMCA estimators, focusing on their biases and suitability for different data characteristics.

Findings

No single estimator is best for all scenarios.

Method choice depends on the dynamic properties of the data.

Each method has specific strengths and limitations.

Abstract

We study finite sample properties of estimators of power-law cross-correlations -- detrended cross-correlation analysis (DCCA), height cross-correlation analysis (HXA) and detrending moving-average cross-correlation analysis (DMCA) -- with a special focus on short-term memory bias as well as power-law coherency. Presented broad Monte Carlo simulation study focuses on different time series lengths, specific methods' parameter setting, and memory strength. We find that each method is best suited for different time series dynamics so that there is no clear winner between the three. The method selection should be then made based on observed dynamic properties of the analyzed series.