Summary
International Symposium on Nonlinear Theory and Its Applications
2016
Session Number:A3L-C
Session:
Number:A3L-C-2
A New Approach to Non-Uniformed Sampled Time Series using Ordinal Partitions
Thomas Stemler, Konstantinos Sakellariou, Michael McCullough, Ben Luo,
pp.-
Publication Date:2016/11/27
Online ISSN:2188-5079
Summary:
The ordinal patterns symbolic methodology may be used to discriminate between deterministic and stochastic time series. Deterministic systems are characterised by forbidden patterns which would not occur even in infinitely long time series, whereas random systems may be exhibit any possible pattern. In this study, we examine the effects of highly irregularly-sampled time series, such those encountered in paleoclimate measurements, on the estimator of the proportion of forbidden patterns manifested by synthetic time series. We investigate three different sampling regimes and observe a high sensitivity in the results. The degree of asymmetry of the sampling distribution is shown to be a useful heuristic indicator for the reliability of the forbidden patterns estimator and the potential for detecting determinism. Exponentially distributed lead to gross underestimation of the total number of forbidden patterns. On the other hand, sampling in the presence of large chronological gaps can lead to relatively accurate estimates as long as the time series contains sufficiently many densely-sampled areas as well.