Hi Joe,
>If I understand the design correctly, the two are equivalent but in either
>case there is a problem of rank deficiency when trying to model attention
>(since it is present in both of the two conditions). Are there also blocks
>of baseline such as rest, fixation, whatever that are not being
>modeled? If not, the effects of attention are not distinguishable because
>there is no experimental variation.
Sorry for not being clear. The stimuli are on for 20s, off for 20s. So
there is always a baseline period between A and B where the subject simply
fixates.
>Another way to think about it is if there are only two blocked conditions,
>you only need one binary variable to model them (ie 1 EV).
So given this stimulus:
A -- baseline -- A & B -- baseline -- A -- baseline -- A & B -- etc
my understanding is that A and B are in fact orthogonal. A voxel that
responds to both A stimuli and B stimuli should have a larger amplitude
during the A & B block. I think one problem here is that it is impossible
to tell if there is an A/B interaction, since we never see B by itself.
But the other problem is the interpretation of the "A&B" regressor in
Option #2. What does that activation mean if it has two cognitive factors
responsible for the activation? In Option #1, the meaning of the
regressors, I think, is clear.
thanks,
jack
|