[Mne_analysis] eegReject option in subject.cov file

[Andres Felipe Salazar Gomez]

Hi MNE users,

Maybe is a very simple issue but I have a question regarding the values of
the --eegReject option in the file used for covariance matrix computation
referred in the MNE user guide as <subject>.cov.

I have been using the 40e-6 value for eegReject (it is used in the sample
data and in the user guide) but this value becomes useless as my subject
goes to other levels. I explain below.

My question is: up to what values is the --eegReject option recommended,
70e-6, 80e-6? How many epochs do you recommend I should get for the noise
covariance matrix?

I am working with a set of "subjects" which are actually the same subject
but at different levels of a "sleep like stages". This "sleep like stages"
affect the subject eeg signals changing the peak-to-peak amplitude of the
epochs (they get bigger). For each "sleep like stage" the peak-to-peak
amplitude is similar so I think the eegReject value is consistent for each
level. Nevertheless, the eegReject value that I used at the first level is
useless at the last levels; I get a good number of epochs for the
covariance matrix in the first level (~179) but for the last ones I do not
even get one and MNE prompts that there is not enough data to calculate
the covariance matrix. At one of the stages that give me a big
peak-to-peak values I have tried with 70e-6 getting only 70 epochs
accepted.

I write the question again: Up to what values is the --eegReject option
recommended in the file used for covariance matrix computation? 70e-6,
80e-6? How many epochs do you recommend I should get for the noise
covariance matrix?

I sincerely appreciate any help in this matter.

-- 
Andres F. Salazar
Research Technologist
Neuroscience Statistics Research Laboratory
asalazar at nmr.mgh.harvard.edu
149 13th Street, Room 4005
Charlestown, MA  02129