EFEL assumes positive amplitude spike peaks and fails to return if spike peaks below 0mV. Druckman and Allen feature extraction protocols seem to tolerate spike peaks below 0mV however.
Option 1. One possibility is to offset membrane potential recordings by a constant, that would boost spike peaks above 0mV. The main danger in performing this action is that it will probably bias and alter a lot of EFEL features. Some EFEL features may not be altered. Under this contingency I suggest setting a flag visible in the pandas data frame to indicate when samples are with regards to spike peaks less than 0mV.
Option 2. Another possibility is to simply not perform EFEL feature extraction on such waveforms and to return None, leading to an impute by 0 condition.
I am going with Option 1, since it seems potentially more information preserving. Imputing by 0 is probably a greater source of bias.
EFEL assumes positive amplitude spike peaks and fails to return if spike peaks below 0mV. Druckman and Allen feature extraction protocols seem to tolerate spike peaks below 0mV however.
Option 1. One possibility is to offset membrane potential recordings by a constant, that would boost spike peaks above 0mV. The main danger in performing this action is that it will probably bias and alter a lot of EFEL features. Some EFEL features may not be altered. Under this contingency I suggest setting a flag visible in the pandas data frame to indicate when samples are with regards to spike peaks less than 0mV.
Option 2. Another possibility is to simply not perform EFEL feature extraction on such waveforms and to return None, leading to an impute by 0 condition.
I am going with Option 1, since it seems potentially more information preserving. Imputing by 0 is probably a greater source of bias.