Predictability in heartbeat data

Predicting the behavior of chaotic dynamical systems is difficult in general. It is important to study such systems since the existence of chaos implies potential short term predictability. Several methods exist to analyze time series, including correlation dimension and the Brock-Dechert-Scheinkman-LeBaron (BDSL) test. Recently, a new tool, sample entropy (SampEn), has gained importance for data differentiation. We have applied these methods to cardiovascular time series data. Our findings suggest that correlation dimension is useful in analyzing such data, but not of sufficient power to discriminate between various data generating processes while sample entropy can be used as a supplementary tool.