Abstract
Heart rate variability (HRV) data display nonstationary characteristics, exhibit long-range correlations (memory) and instantaneous variability (volatility). Recently, we have proposed fractionally integrated autoregressive moving average (ARFIMA) models for a parametric alternative to the widely-used technique detrended fluctuation analysis, for long memory estimation in HRV. Usually, the volatility in HRV studies is assessed by recursive least squares. In this work, we propose an alternative approach based on ARFIMA models with generalized autoregressive conditionally heteroscedastic (GARCH) innovations. ARFIMA-GARCH models, combined with selective adaptive segmentation, may be used to capture and remove long-range correlation and estimate the conditional volatility in 24 hour HRV recordings. The ARFIMA-GARCH approach is applied to 24 hour HRV recordings from the Noltisalis database allowing to discriminate between the different groups.