Classification result by dropping some side electrodes

External Email - Use Caution

Hi all,
I have eeg data for hand movement recorded from 32 electrodes. I did the
classification and got an accuracy. I removed the electrodes on the sides.
And did the classification for only 16 electrodes (over the motor cortex).
I got lower accuracy. (I think this is due to the high connectivity).
Then I wanted to see how the classification would be for the rest
electrodes. So I dropped (removed) the electrodes over the motor cortex and
chose only the electrodes on the sides. Surprisingly I got better results
by choosing the electrodes on the sides than the case when choosing the
electrodes over the motor cortex?
Is it normal? I'm new to this. I think you have more experience and
knowledge than me.

Thanks in advance for any help
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190427/a5d468c6/attachment.html

External Email - Use Caution

hi,

how many trials/samples do you have? What is the variance of the accuracy
over folds? I am asking to know if your conclusions of "better" are strong
or could be due to variance/randomness in the estimation.

Alex

External Email - Use Caution

Thank you very much for the answer. I have 35 samples from first class vs
65 samples from second class. I'm using the TSClassifier with the
covariance matrices. The variance in the accuracy was about 4%.

Thanks for your help and for your suggestions

Alexandre Gramfort <alexandre.gramfort at inria.fr>, 30 Nis 2019 Sal, 09:57
tarihinde ?unu yazd?:

        External Email - Use Caution

hi,

how many trials/samples do you have? What is the variance of the accuracy
over folds? I am asking to know if your conclusions of "better" are strong
or could be due to variance/randomness in the estimation.

Alex

>
> External Email - Use Caution
>
> Hi all,
> I have eeg data for hand movement recorded from 32 electrodes. I did the
classification and got an accuracy. I removed the electrodes on the sides.
And did the classification for only 16 electrodes (over the motor cortex).
I got lower accuracy. (I think this is due to the high connectivity).
> Then I wanted to see how the classification would be for the rest
electrodes. So I dropped (removed) the electrodes over the motor cortex and
chose only the electrodes on the sides. Surprisingly I got better results
by choosing the electrodes on the sides than the case when choosing the
electrodes over the motor cortex?
> Is it normal? I'm new to this. I think you have more experience and
knowledge than me.
>
> Thanks in advance for any help
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> Mne_analysis Info Page

_______________________________________________
Mne_analysis mailing list
Mne_analysis at nmr.mgh.harvard.edu
Mne_analysis Info Page

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190430/6db9c37f/attachment-0001.html