[Eeglablist] t-test/compare ERPs

[Arnaud Delorme]

Gopakumar Venugopalan wrote:
> Greetings Arno, before I get to the EEGLAB question, wanted to mention 
> that I knew someone (Denise) at the Univ of GA with a  DeLorme last 
> name--any relation?
>
> I have read the very well laid tutorial several times, and can 
> understand it very well. Please bear with me when I ask:
> i) I have epoched data (-150-0-3600ms long with 954 data points). These 
> epochs or averages were obtained using 4D software where continous data 
> was averaged to two triggers. They produced two files: one for the word 
> condition and, one for the non-word condition. I have 12 separate files 
> for each condition.
> ii) The MEG data was imported using the import function (4d 
> neuroimaging option) and saved as *.set files.
> iii) EEG data that was concurrently obtained was saved separately and 
> also imported (neuroscan.eeg option) from neuroscan.avg files and also 
> saved as a separate set of *.set files.
> iv) I also installed grandaverage 0.9 (Widmann) and it seems to work 
> well from the command line.
> v) I intend to direcly do the paired t-test between the two 
> conditions for the two modalities(or sets of files) separately
> vi) would you be kind enough to tell me what can be done and the 
> most expeditious route. Something as simple as Start>create study> so 
> on would be fine. I saw something tutorial screen where there were 
> integers in a box for condition one 1 2 3 and 4 5 6 in box for 
> condition two. I was expecting to see individual files--in my mind!!
> The only reason I ask this is I have circumvented the preprocessing 
> step due to logistical reasons, and therefore have epoched data to work 
> with.
>   

The fastest way is to use menu item "Plot > Sum/compare ERPs". When you 
have all your dataset loaded (and assuming you have enough space!) 
simply enter the index of the datasets for the two conditions. This 
function will disappear though because the solution below is more robust.

The safest and most reliable route is to use STUDY. Enter each dataset 
in the STUDY with the name of the condition. Then precompute the ERPs 
(you will also be able to subtract artifactual components if any in the 
next EEGLAB reversion). Then plot ERPs and compute statistics. The 
tutorial for STUDY explains each step in detail.

http://sccn.ucsd.edu/eeglab/clusttut/clustertut.html

Hope this helps,

Arno