Just noticed from the gitbook, the illustration of the high_fluctuation (i.e., MD_hrv_classic_pnn40) is that "it computes the proportion of difference magnitudes that are greater than 4% of the standard deviation of the time series."
It contradicts its setting in the code base, where it looks more like computing the proportion of difference magnitudes that are greater than 4% of a scaling factor \sigma, where it was chosen to be 1000.
Statements from the original paper of those pnn features might favor the scaling factor possibility more. Thoughts and discussion are welcome!
Just noticed from the gitbook, the illustration of the high_fluctuation (i.e., MD_hrv_classic_pnn40) is that "it computes the proportion of difference magnitudes that are greater than 4% of the standard deviation of the time series."
It contradicts its setting in the code base, where it looks more like computing the proportion of difference magnitudes that are greater than 4% of a scaling factor \sigma, where it was chosen to be 1000.
Statements from the original paper of those pnn features might favor the scaling factor possibility more. Thoughts and discussion are welcome!