[Eeglablist] two ICA analysis questions

[Siri-Maria Kamp]

Hello EEGlab users,

I have run an ICA (infomax) on EEG data from an experiment with two
conditions. I have two questions that came up during that analysis.

1) I have used the function compvar() to get the variance of the original
data (the data I ran the ICA on) accounted for by each component. The
second component turned out to account for less variance than the third
and fourth, respectively. As runica() sorts the components in descending
order of variance accounted for, should compvar() not also return
variances that decrease for increasing component numbers? I am trying to
figure out if I just understood something wrong, or if I use the function
compvar() incorrectly.

2) One can use comperp() to conduct a t-test for every time point that
compares the IC activations in the two conditions. Is it legitimate to
alternatively use the maximum (or minimum) activation of an IC within a
given time window, similar to peak picking methods in ERP analyses, and
perform a ttest on these activations?

Thank you very much for your answer, I appreciate your help!
Siri



-- 
Siri-Maria Kamp, BSc
Graduate Student
USF EEG laboratory
Department of Psychology
University of South Florida