Information-based detection of nonlinear Granger causality in multivariate processes via a nonuniform embedding...

Area of Science:

• Dynamical Systems Analysis
• Information Theory
• Causality Assessment

Background:

• Assessing nonlinear causality in complex systems is challenging.
• Existing methods often struggle with multivariate time series.
• Understanding system interactions requires robust causality detection.

Purpose of the Study:

• To develop an information theory-based approach for nonlinear causality.
• To introduce a sequential, nonuniform embedding procedure for time series.
• To quantify causal coupling using information transfer.

Main Methods:

• Sequential nonuniform embedding of multivariate time series.
• Minimization criterion based on conditional entropy.
• Corrected conditional entropy estimator for quantization bias.
• Granger causality for predictability improvement detection.

Main Results:

• The proposed method demonstrates superiority over standard uniform embedding in simulations.
• It effectively detects causal coupling and quantifies information transfer.
• Investigated effects of quantization, data length, and noise contamination.

Conclusions:

• The developed approach provides a robust framework for nonlinear causality assessment.
• Applicable to both deterministic and stochastic dynamical systems.
• Successfully applied to cardiovascular regulation and brain activity analysis.