An empirical evaluation of alternative methods of estimation for Permutation Entropy in time series with tied values

Bandt and Pompe introduced Permutation Entropy in 2002 for Time Series where equal values, xt1 = xt2, t1 = t2, were neglected and only inequalities between the xt were considered. Since then, this measure has been modified and extended, in particular in cases when the amount of equal values in the series can not be neglected, (i.e. heart rate variability (HRV) time series). We review the different existing methodologies that treats this subject by classifying them according to their different strategies. In addition, a novel Bayesian Missing Data Imputation is presented that proves to outperform the existing methodologies that deals with type of time series. All this facts are illustrated by simulations and also by distinguishing patients suffering from Congestive Heart Failure from a (healthy) control group using HRV time series