Conditional volatility properties of sleep-disordered breathing.

The present paper investigates the conditional volatility properties of apneic electrocardiogram (ECG) sequences taken from the Physionet apnea database using a set of GARCH models. We discuss the theoretical justification of this approach in the light of the recent developments in heart rate volatility studies. The findings support the value of the Integrated GARCH(1,1) specification in fitting apneic ECG segments implying the presence of shocks to conditional volatility that persist over time. The ARCH(1) model was found superior in fitting normal ECG recordings. These findings provide an important indication to future research that GARCH models can be employed in the development of an apnea screening tool to be used in clinical settings.