Mne_analysis Digest, Vol 81, Issue 9

✉️ Mailing List Archive (read-only)
1

Thanks you for all the answers and precisions,

I'm going to follow Denis' suggestions and try to use baseline data as
material for computing covariance matrices, as it ensure hat all treatments
applied to the data are also applied to baseline. As my baseline is only
200ms, i'm going to try regularization. For the particulary artifacted
subjects, I'm going to re-examine carefully the raw data, as I could have
forgot or misrecord bad channels.

Just one last question: what is the gain of combining all sensor types
(grads, mags, eeg) ? Here I know from visual inspection and sensor-level
analysis that my eeg data is noisier than meg ...

thanks again!

Baptiste

2014-10-02 8:54 GMT+02:00 <mne_analysis-request at nmr.mgh.harvard.edu>:

2

Just one last question: what is the gain of combining all sensor types
(grads, mags, eeg) ? Here I know from visual inspection and sensor-level
analysis that my eeg data is noisier than meg ...

it's always a good idea to try to combine them all unless EEG looks crap
or EEG forward is bad due to bad bem + segmentation.
To check try localizing something simple with EEG only.

HTH
Alex

3

MEG and EEG provide complementary measures of the electromagnetic
field/potential, so combining them should improve localization. For
example, see this paper by Sharon et al.:

http://www.ncbi.nlm.nih.gov/pubmed/17532230

Eric