Session S33.4

Long Memory and Volatility in HRV: An ARFIMA-GARCH Approach

A Leite, AP Rocha*, ME Silva

Porto University
Porto, Portugal

Heart Rate Variability (HRV) data display non-stationary characteristics, exhibit long-range correlations and instantaneous variability (volatility). Recently, we have proposed the use of ARFIMA (Fractionally Integrated Autoregressive Moving Average) models as an alternative approach to the widely-used technique DFA, 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 GARCH (Generalized Autoregressive Conditionally Heteroscedastic) innovations. ARFIMA-GARCH models, combined with selective adaptive segmentation, may be used to capture and remove long-range correlation and estimate the conditional volatility, leading to an improved description of the components in 24 hour HRV recordings. This modelling can be used in reduced length segments of 512 beats.
The ARFIMA-GARCH approach is applied to 24 hour HRV recordings of 30 subjects from the Noltisalis data base: 10 Healthy subjects, 10 patients suffering from Congestive Heart Failure (CHF) and 10 heart Transplanted patients. The results (mean +/- standard deviation) for the long-memory parameter (d) during the 24 hours (P1), 6 hours of night (P2) and day (P3) periods, are for each group: Healthy (0.44 +/- 0.06, P1; 0.34 +/- 0.07, P2; 0.46 +/-0.09, P3); CHF (0.52 +/- 0.14, P1; 0.38 +/- 0.16, P2; 0.59 +/- 0.16, P3); Transplanted (0.76 +/- 0.10, P1; 0.67 +/-0.17, P2; 0.78 +/- 0.12, P3), respectively. The corresponding results for the conditional volatility parameter (u) are: Healthy (0.23 +/- 0.09, P1; 0.20 +/-0.04, P2; 0.24 +/- 0.14, P3); CHF (0.15 +/- 0.08, P1; 0.16 +/- 0.08, P2; 0.12 +/- 0.05, P3); Transplanted (0.11 +/-0.06, P1; 0.11 +/- 0.07, P2; 0.10 +/- 0.07, P3). The results indicate that the long memory parameter has a circadian variation, with different regimes for night and day periods. Moreover, increased long memory values and decreased conditional volatility values for sick subjects, suggest that ARFIMA-GARCH modelling allows discrimination between the different groups.

(Abstract Control Number: 140)