[Eeglablist] Coherence with few trials

[Andersen S.B.]

Hi
 
We have EEG data from a social psychophysiological experiment, where we
systematically manipulated nonverbal cues adopted by a live confederate in
order to examine EEG personality correlates of sensitivity to real
interpersonal distance (2, 4, 8 and 16 feet), gaze (averted, direct), and
smile.
 
Participants were seated in an elevated adjustable chair, ensuring that the
participant and confederate were in same eye level. We had a 9 seconds
exposure time where the participant was instructed to gaze into the
confederate's eyes, irrespective of the confederate's nonverbal behaviour,
which was followed by a 9 seconds intertrial period where the participants
closed their eyes and verbally rated the pleasantness of the previous trial.
There were 12 unique trials (4 distances X 3 facial behaviours), each
repeated 6 times.
 
One of the problems with the design is that we could not control how quickly
the participants opened their eyes, so the first second of the trial had to
be discarded, leaving 8 seconds for the trial (and with 6 repeats, a total
of 48 seconds per condition).  Of course, this varied slightly from
participant to participant, some took 200 ms to start opening their eyes,
others 400 etc.  Thus, there was not a strict time locking of the EEG record
and the stimulus signal (i.e., the confederate).
 
We now wish to apply coherence analysis to some of the data using EEGLAB if
possible, but we've run into a few problems so any advice on the matter
would be greatly appreciated. 
 
1.             Length of epoch
 
Since each of the 12 unique trials lasts 8 seconds, and is repeated 6 times
(a total of 48 seconds), what is the best way to epoch my data for coherence
analysis?  Should I divide each 8 second trial into say epoch of 2 seconds,
yielding a total of 24 epochs per condition? Or can I leave the epoch length
to 8 seconds, which would only amount to 6 trials per condition for the
coherence analysis? Since I am not yet that knowledgeable about coherence
analysis, I don't know whether there is a minimum number of epochs needed
for a coherence analysis. 
 
2.             baseline removal
Given that the EEG data are not exactly timelocked to the "confederate's
behaviour", is there any other way of removing baseline activity?  Is it
best to define the first second as baseline and the other 7 seconds as trial
data?  
 
Any advice on the matter would be greatly appreciated.
 
Regards
 
Soren Andersen
 
 
Dr. Soren Bo Andersen
Department of Psychology
University of Wales Swansea
Singleton Park
Swansea SA2 8PP
 
tel:  + 44 (0) 1792 295713
fax: + 44 (0) 1792 295679
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20050623/3d3d997b/attachment.html>