Modeling volatility in heat rate variability.

Area of Science:

• Physiology
• Biomedical Engineering
• Data Science

Background:

• Heart Rate Variability (HRV) data analysis is crucial for clinical applications and research.
• Traditional methods using short-memory AutoRegressive (AR) models with recursive least squares struggle with HRV's long memory and time-varying volatility.
• Existing models do not fully capture the complex characteristics of conditional volatility in HRV data.

Purpose of the Study:

• To develop and apply advanced parametric models for HRV data analysis.
• To effectively model and characterize the long memory and heteroscedasticity (time-varying volatility) present in HRV.
• To improve the understanding and clinical utility of HRV recordings.

Main Methods:

• Utilized long memory Fractionally Integrated AutoRegressive Moving Average (ARFIMA) models.
• Incorporated nonlinear Generalized Autoregressive Conditionally Heteroscedastic (GARCH) and Exponential Generalized Autoregressive Conditionally Heteroscedastic (EGARCH) models to handle heteroscedasticity.
• Applied ARFIMA-EGARCH models to 24-hour HRV recordings from the Noltisalis database.

Main Results:

• The ARFIMA-EGARCH models successfully captured and removed long memory from the HRV data.
• Characterized the conditional volatility in 24-hour HRV recordings, accounting for clustering and asymmetry (leverage effects).
• Demonstrated the suitability of these advanced models for complex HRV data analysis.

Conclusions:

• ARFIMA-EGARCH models provide a robust framework for analyzing HRV data with long memory and volatility.
• These models offer enhanced characterization of HRV compared to traditional methods.
• The findings support the use of these advanced modeling techniques for clinical and research purposes in HRV analysis.