[Eeglablist] improving coherence estimates

[Spencer, Kevin M.]

Dear Stan,

Regarding your first question, you should check out Truccolo et al., Clinical Neurophysiology 2002 113:206-226. In this study they demonstrated that subtraction of the ERP from single trials actually leaves residual activity on the single trials which distorts estimates of single-trial activity. This is highly relevant for both coherence and single-electrode analyses.

While subtracting the ERP from the single trials before subsequent analyses is not a good idea, if you're doing time-frequency analyses, you can subtract evoked power/coherence from total power/coherence (average of single-trial power/coherence).

Kevin
--------------------------------------------------------------------------------------------------
Kevin M. Spencer, Ph.D.
Director, Neural Dynamics Laboratory (http://ndl.hms.harvard.edu<http://ndl.hms.harvard.edu/>)
Research Health Scientist, VA Boston Healthcare System
Assistant Professor of Psychiatry, Harvard Medical School
--------------------------------------------------------------------------------------------------

________________________________
From: eeglablist-bounces at sccn.ucsd.edu [eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Stanley Klein [sklein at berkeley.edu]
Sent: Saturday, May 29, 2010 5:40 AM
To: eeglablist at sccn.ucsd.edu
Subject: [Eeglablist] improving coherence estimates

Dear EEGlab,
   In three days we'll be presenting a poster at a brain-computer
interface (BCI) conference in Monterey, CA on the topic of how to
reduce noise in coherence estimates for BCI purposes. I would
like to ask EEGlab whether the two approaches of the poster have
already been published or discussed. I'm not familiar with prior
work on it and I would really like to know before I make a fool of
myself.

1) The usefulness of removing the ERP before calculating coherence.
Our simulations are the main topic of the poster.

2) The usefulness of using Cauchy wavelets as filters for time-frequency
analysis when one wants high resolution in time (like 1 to 1.5 cycles).
These filters have rapid falloff at low temporal frequencies so they are
appropriate for the 1/f nature of EEG noise.

Again, I'd be grateful for any leads to articles on either of this items.
And I look forward to seeing some of you in Monterey.

thanks,
Stan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20100601/a9656897/attachment.html>