Unbiased entropy estimates in stress: a parameter study.

This study proposes a method for obtaining a stable working point that enables the unbiased estimates of the approximate entropy and the sample entropy. Pattern length, normalized threshold, time delay and tines series length are levels of freedom that are considered. Pulse interval signals used for the experiment are recorded from laboratory animals with different likelihood of developing cardiovascular disease--normal and borderline hypertensive rats--exposed to the acute and to the chronic stress. It is shown that the threshold level is the major source of the instability, and that the generally accepted and widely used methods for a threshold choice may lead to an incorrect psychological interpretation. A method for the threshold level correction is proposed.