[Eeglablist] sleep study by analyzing all-night EEG

[arno]

Dear Masaki,

> 1. How could I get the matrix data of spectra alalysis of each channel
> and frequency?
> I want to compare the magnitudes of frequency band among the subjects,
> so it would be useful to extract data as matrix in the workspace. For
> long time I have used a FFT method, so I'm not familiar with
> multitaper methods.
The spectopo function returns the spectrum for all channel for one
subject. You have to run it on each subject and then average all these
spectra. This is using the Welch method (repeated FFT). For
time-frequency decomposition there is a script example below

http://www.sccn.ucsd.edu/eeglab/scripttut/script_tutorial.html#script%20for%20performing%20time-frequency%20decompositions

> 2. In sleep study, long-lasting period (for example, slow wave sleep
> or REM sleep) is our interest. Is it possible to extract such a long
> duration as epochs? If impossible, I think to use data after rejecting
> other sections.
No, all epochs must have the same length. A future (to be released soon)
version of timef() can though use multiple baseline at different
latencies (baseline can be the sleep period for instance).
> 3. I would like to know whether the ICA is useful for evaluating sleep
> recording. We can find various forms of waves in the human sleep EEG,
> such as big delta wave during non-REM sleep, low scattered fast wave
> during REM sleep. I wonder if we can use ICA or
> other rejecting artifacts events.
Yes, it is useful to reject artifact for sure. About studying brain
dynamics, I cannot see why not. I have not tried myself though.

Best,

Arno