[Eeglablist] Re: grand averaging

Arnaud Delorme arno at salk.edu
Wed Mar 16 16:15:09 PST 2005


Dear Patti,

>I have data for 23 subjects (out of a projected 50) for a task with two 
>conditions.  What would be the most effective way to look at the data to 
>determine trends/differences within subject and across conditions, etc.:
>  
>
The most efficient way to compute grand-ERP etc... is probably not to 
use EEGLAB Graphic interface since for now you will need to have all the 
data from all your subjects in memory. In the near future, EEGLAB 
Graphic interface will allow you to leave the data on the disk while 
still accessing dataset information, so this will not be a problem anymore.

>1.  average the segmented eeg data for each subject and then use the "sum/
>compare ERPs" function?
>  
>
If you want to use this function from the GUI, you should average all 
trials after reading each dataset as shown below. This way, you do not 
take the risk of saturating the memory and the "sum/compare ERPs" 
function will return the same result as if you had the raw data trials.

EEG.data = mean(EEG.data,3); % average
[ALLEEG EEG CURRENTSET] = eeg_store(ALLEEG, EEG, CURRENTSET); % store

>2.  average the segmented eeg data for each subject AFTER running the ICA 
>function for each subject and then "sum/compare"?
>  
>
You need to regroup components that are potentially the same across 
subject before averaging their activities. Clustering features will be 
available soon in EEGLAB but for now you have to do it on the command 
line (by hand or using a clustering algorithm).

>3.  Before averaging, do I need to, for example, transfer the data to BESA to 
>get a standard 81 electrode file/set-up, so that I have a standard format in 
>order to compare/average different individuals (as their electrode locations 
>vary according to size of net used/size and shape of head)? 
>  
>
No, BESA electrode location files can be automatically read in EEGLAB 
(see tutorial below).

http://sccn.ucsd.edu/eeglab/maintut/channel_locations.html

Arno




More information about the eeglablist mailing list