[Eeglablist] Analyzing ERSPs

[Guillaume Rousselet]

Hey Bornali,

You could do different analyses to address different questions.

[1] Whatever you do it is a good idea to use cluster based statistics,  
with bootstrap or permutation.

[2] If you're main interest is to show the reliability of the ERSP  
responses within subject, you can run single-trial analyses to compare  
session 1 and session 2 in each subject. Reliability should result in  
small if any differences. You may want to centre your data first to  
remove mean session effects if you are interested in the reliability  
of the effects of different conditions. If your design includes  
different conditions, you could run, at each time and each electrode,  
an ANOVA including a session factor, and correct for multiple  
comparisons using a cluster test.

[3] If you're main interest is to show the reliability of the ERSP  
responses across subjects, you can run group comparisons of the two  
sessions, using the mean ERSP from each subject (thus after removing  
the single-trial variance associated with each subject).

[4] You could perform [3] after fitting a model on each subject, and  
perform a pairwise group comparisons on the beta coefficients of the  
model, similarly to what people do in fMRI. This would tell you if you  
have significant session effect across subjects. Of course this  
analysis will not reveal session effects that are not consistent in  
space and time across subjects - the typical problem with group  
statistics.

Best wishes,

Guillaume



On 7 Jul 2010, at 20:28, Bornali Kundu wrote:

> Hello,
>
> This question is not directly related to using EEGLAB, but has to do  
> with analyzing ERSPs.  What is an acceptable way to statistically  
> compare ERSPs within subjects over multiple sessions?  I have 10  
> subjects, where each repeated the same task during 2 separate  
> sessions.  I want to compare ERSP amplitude over multiple sessions  
> for each subject with ERSP amplitude over many subjects within one  
> session.  Does it make sense to run some type of ANOVA at every time  
> and frequency point or a MANOVA?  Or is there a better way to  
> compare two ERSPs taking into consideration the "pattern" of  
> activity over time and frequency (perhaps using a cluster-based  
> permutation strategy)?  Any thoughts would be greatly appreciated.
>
> Thanks!
>
> Bornali
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu


************************************************************************************
Guillaume A. Rousselet, Ph.D., lecturer

School of Psychology
Institute of Neuroscience & Psychology
Centre for Cognitive Neuroimaging (CCNi)

The University of Glasgow, charity number SC004401

http://www.psy.gla.ac.uk/staff/index.php?id=GAR01

Email: g.rousselet at psy.gla.ac.uk
Fax. +44 (0)141 330 4606
Tel. +44 (0)141 330 6652
Cell +44 (0)791 779 7833


"Corporations have only one duty: to promote their own and their  
owners' interests. They have no capacity, and their executives no  
authority, to act out of a genuine sense of responsibility to society,  
to avoid causing harm to people and the environment, or to work to  
advance the public good in ways that are unrelated to their own self- 
interest. Deregulation thus rests upon the suspect premise that  
corporations will respect social and environmental interests without  
being compelled by government to do so. No one would seriously suggest  
that individuals should regulate themselves, that laws against murder,  
assault, and theft are unnecessary because people are socially  
responsible. Yet oddly, we are asked to believe that corporate persons  
- institutional psychopaths who lack any sense of moral conviction and  
who have the power and motivation to cause harm and devastation in the  
world - should be left free to govern themselves."

	Joel Bakan - The Corporation - 2004
************************************************************************************



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20100708/cf4ab809/attachment.html>