I have a problem with analyzing my EEG power spectral density.
you know when we compute the EEG power spectral density we have a graph like the below picture
always when I compute the EEG PSD, delta (0-4) and theta (4-8) bands have high values and graphs come from high to low.
Actually, in my work, I'm looking for which EEG bands activate during my experiment but always delta and theta are greater than others. How can I trust these 2 bands are really more activated during my task?
just in 1 PSD, I can recognize the Alpha band is activated.
Hello Hamad,
As you have realized, EEG has a 1/f-like power spectrum. Therefore it is usually recommended that you normalize your data. Depending on the dataset and the analysis you are doing you can normalize with a 1/f normalization (spectral flattening), baseline normalization (according to a non-task baseline) or comparing two tasks (subtracting to get the difference in task power only).
Hope this helps.
Beth
and I have an another question about the difference between pwelch and pburg for computing PSD.
For example, I used these 2 codes for computing PSD of 1 channel of my signal.
If you are specifically comparing two conditions, you should compute a contrast between the results you get between the two conditions. Either computing a different (for one subject) or a t-test (group study).
If you prefer this other method for evaluating the spectrum, use it. You should however try first to adjust the parameters of the pwelch method (use shorter estimator windows).
Just remember you should not be selecting the tools you use in your analysis based on the fact they give the results you want, but based on some reference literature.
