Considerations for Applying Entropy Methods to Temporally Correlated Stochastic Datasets

Area of Science:

• Biomechanical Analysis
• Physiological Signal Processing
• Stochastic Data Modeling

Background:

• Biomechanical and physiological variables often exhibit temporal correlations.
• Sample Entropy (SampEn) is a common method for analyzing signal regularity.
• Applying SampEn to correlated data presents analytical challenges.

Purpose of the Study:

• To highlight considerations and provide recommendations for applying Sample Entropy (SampEn) to temporally correlated stochastic datasets.
• To address analytical issues in biomechanical and physiological data analysis.
• To improve the reliability of entropy-based measures in complex datasets.

Main Methods:

• Simulated temporally correlated data using Autoregressive Fractionally Integrated Moving Averaged (ARFIMA) models.
• Applied ARFIMA modeling to quantify temporal correlations and classify data stationarity.
• Utilized SampEn to assess dataset regularity and evaluated data cleaning strategies.

Main Results:

• ARFIMA modeling effectively estimates temporal correlation properties and classifies datasets.
• ARFIMA-assisted data cleaning mitigates outlier influence on SampEn estimates.
• SampEn has limitations in distinguishing between different stochastic datasets; parameter normalization is ineffective for stochastic data.

Conclusions:

• ARFIMA modeling is a valuable tool for analyzing correlated biomechanical data and enhancing SampEn reliability.
• Complementary measures are suggested for a more comprehensive characterization of biomechanical variable dynamics.
• The study provides practical recommendations for researchers applying entropy methods to complex physiological signals.