The most important characteristic of a covariance stationary self-similar stochastic process is that it is long-range dependent. The long-range dependent time series hold significant correlations across arbitrarily large time scales. And the Hurst parameter H measure the degree of long-range dependence and can be estimated by several methods.
Great peace of code! I was looking long time around to find it.
- Maybe the various methods could briefly be summarized so that terminological variations can be cleared.
I would like to estimate the Hurst coefficient so I can determine if my data series is persistent (H>0.5), antipersistent (H<0.5), or neutral (H=0.5). Several of the estimates I get are, for example, H = 1.8, or H=1.6, H=0.9....should I disregard the whole number and concentrate only on the decimals for interpreting persistence?