[Eeglablist] Simultaneous two-device recording: normalization needed?

[Stefan Dvoretskii]

Dear list members,

I have used two same devices (Mentalab Explore™) to record the data with
two 8-channel devices. My pipeline consists of normalisation of the data
using (data-1/2*(reference channel)), where reference channel is placed on
the mastoid. Then, I perform LDA with variance, RMS and PSD of the beta and
mi frequency bands of separate channels as features.
Interestingly enough, in this setup I get a reasonable accuracy for each
one of the devices, but a lower accuracy when combining both devices' data
in one EEG structure. Hence the question: has anyone tried the same setup
with two devices? Any tips on what could be the reason for such low
prediction accuracy with LDA? Maybe I have to co-normalize the devices data
in some way?

Best regards,
Stefan