[Eeglablist] factorial repeated measures of ANOVA

[shirazi at ieee.org]

Hi,
I need some help with setting up factorial repeated measures of ANOVA for
EEG source-level measures. The model I am looking for is similar to the
following (hypothetical) scenario:

1- Suppose that we have four arm-reaching visuomotor perturbation
experiments, 2 with the right hand and 2 with the left hand. All of the (n)
subjects complete all four experiments. Looking at the ERSP image of a
cluster at the SMA, there is desynchronization in the beta band shortly
after the perturbation, with the right side experiments tend to create
stronger desynchronization. To test the sidedness hypothesis, we average
beta-band power after the perturbation for all the experiments (e=4) and
subjects (n). For now, we don't look at the single trials, so each subject
has only one value for the power for each experiment. Therefore, we will
have a stats = [e x n] matrix with the first two rows represent the right
hand and the second two represent the left hand experiments.
I tried using "statcond" with data = {stats(1:2, 1:n), stats(3:4, 1:n) }.
However the code runs with a paired T-test assumption. Even forcing ANOVA,
it returns two p-values instead of one. Upon more investigation, if I add
more experiments to the left or right side, say 3 right and left side
experiments (hence, data = { stats(1:3, 1:n), stats(4:6, 1:n) }), then I
would get three p-values which apparently does not seem to be repeated
measures at all.

If the scenario above has a solution, I wonder if I can test multiple
factors as well: with the same scenario, suppose that the experiments on
each right or left side are different, for example, one experiment is
clockwise visuomotor rotation, and the other is counterclockwise rotation.
Therefore, there are two factors: A = side, and B = rotation direction. So
for we have the following stats matrix:
stats(1, n)    A1B1
stats(2, n)    A1B2
stats(3, n)    A2B1
stats(4, n)    A2B2

Then the results should demonstrate the effects of the side, rotation
direction, and interaction of side and rotation direction. Unfortunately, I
could not find such a design in EEGLAB at all. (I think LIMO has some
capability for factorial tests though.)

I am sorry that my question got this long. I wanted to try my best to be as
clear as possible.

Best
Seyed

--
Seyed Yahya Shirazi
GRA, BRaIN Lab
University of Central Florida