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

Area of Science:

• Physical sciences
• Data analysis
• Information theory

Background:

• Local mixing of data points is a common issue in physical experiments.
• This mixing can arise from measurement apparatus or natural processes like diffusion.
• Understudied effects of local mixing can impact data interpretation.

Purpose of the Study:

• To propose a model-free technique for detecting local mixing in time-series data.
• To quantify the scale at which data mixing occurs.
• To provide a tool for scientists to set appropriate data measurement scales.

Main Methods:

• Utilized permutation entropy, an information-theoretic technique.
• Analyzed patterns by varying the temporal resolution of the calculation.
• Validated the method on synthetic datasets and real-world experimental data.

Main Results:

• Successfully detected local mixing in time-series data across various applications.
• Quantified the scale of mixing, providing insights into data reliability.
• Demonstrated the technique's effectiveness on chemistry experiments, methane records, and ice core data.

Conclusions:

• Permutation entropy offers a robust, model-free approach to identify and scale local data mixing.
• This method enhances the reliability of scientific data by informing measurement scale selection.
• The technique is broadly applicable to diverse scientific fields dealing with time-series data.