[Eeglablist] Problem with Envtopo

nicole strang nicole_strang at hotmail.com
Mon Jun 13 09:55:28 PDT 2005


Hi Everyone,
I have an ERP dataset, collected using the GSN128 montage, from 16 subjects 
in an experiment with two conditions. We initially analyzed the data by 
comparing the grand averages for each condition using Netstation tools. We 
found focal differences using the bootstapping method and wanted to further 
elucidate the scalp distribution of the generator(s) that accounted for 
theses differences. To do we concatenated the single subject averages for 
each condition (32 total) and imported them into EEGLAB. We also reduced the 
number of electrodes from 128 to 67 to provide sufficient data for ICA, 
which we did using runica. Once we conducted the ICA we created a new 
dataset (comprised only of the control condition, 16 epochs) from the parent 
dataset. We used the EEGLAB option 'component ERPS with component maps 
(compare)' to subtract the control condition from the parent dataset and 
display the 7 largest components in the timeframe from 300ms to 400ms. 
post-stimulus (this is the time period where we for ERP condition 
differences). We did this with the thought that the components accounting 
for the greatest variance would account for condition differences. My first 
questions is whether this is an appropriate method of analysis? Secondly, 
according to the latest eeglab manual there sould be a printout of the 
variance accounted for by each component, however this is not the case. Is 
there any way I can determine how much variance is accounted for by each 
component? I realze that when you select each topoplot you get rv-Is this 
relative variance or residual variance? Finally, I have a general ICA 
question. In my analyses I get rvs and pvafs (I used a previous version of 
EEGLAB to get the printout of pvaf for each component, however in this 
version I recieve an error when I try to subtract on dataset from the other) 
that are negative. What does this mean?
Thank you very much for your help,
Nicole Strang




More information about the eeglablist mailing list