Sample Entropy Computation on Signals with Missing Values.

Area of Science:

• Time Series Analysis
• Entropy Estimation
• Signal Processing

Background:

• Sample entropy quantifies time series complexity by embedding data into m-dimensional spaces.
• Missing or invalid data points complicate distance calculations in embedding spaces.
• Existing methods like deletion and interpolation have limitations in handling such data.

Purpose of the Study:

• To propose a novel algorithm for computing sample entropy that directly accommodates missing or invalid values.
• To compare the proposed method against deletion and interpolation techniques.

Main Methods:

• The novel algorithm embeds time series into m-dimensional spaces, handling missing values directly within this space.
• Theoretical and experimental comparisons were conducted against deletion and interpolation preprocessing methods.

Main Results:

• The proposed algorithm effectively handles missing or invalid data points in the embedding space.
• It demonstrates significant advantages over deletion and interpolation.
• The new methodology consistently shows the lowest deviation in expected sample entropy values.

Conclusions:

• The novel sample entropy computation method offers a robust solution for time series with missing or invalid data.
• It provides more accurate entropy estimations compared to traditional preprocessing techniques.
• This approach enhances the reliability of time series complexity analysis in the presence of data imperfections.