[Eeglablist] Intuition behind negative but strong correlation among EEG channels

Velu Prabhakar Kumaravel velu.kumaravel at unitn.it
Thu Sep 19 06:39:40 PDT 2019


Dear all,



One of the pre-processing steps used in the "Clean_rawdata" pipeline is to
remove bad channels based on Correlation as the criterion (using channel
location information). I found the following line of code which performs
the threshold comparison and sets a flag vector to 1 or 0 depending on
whether the criterion is met or not.

 flagged = corrs < corr_threshold;

Theoretically, correlation ranges from -1 to +1. But, using the above
implementation, a channel that *strongly but negatively* correlates with
other channels is considered “bad”. I wonder if it instead should look for
the absolute correlation before checking the threshold. However, I’m not
sure if it is appropriate with EEG signals. Any insights on this are
greatly appreciated.



P.s. I work with neonate’s data captured using Geodesic EEG system (GES400
EGI, USA).


Sincerely

Velu Prabhakar Kumaravel


More information about the eeglablist mailing list