Detection of local mixing in time-series data using permutation entropy

While it is tempting in experimental practice to seek as high a data rate possible, oversampling can become an issue if one takes measurements too densely. These effects take many forms, some of which are easy detect: e.g., when the sequence contains multiple copies same measured value. In other situations, there mixing$-$in measurement apparatus and/or system itself$-$oversampling be harder detect. We propose novel, model-free technique detect local mixing time series using information-theoretic called permutation entropy. By varying temporal resolution calculation and analyzing patterns results, we determine whether mixed locally, on what scale. This used by practitioners choose appropriate lower bounds scales at measure or report data. After validating this several synthetic examples, demonstrate its effectiveness from chemistry experiment, methane records Mauna Loa, Antarctic ice core.