Dear Anna,
> we have a study with a problematic fMRI design to deal with. There are
> two conditions A and B, each presented to the subjects in separate
> successive sessions (several minutes break between the sessions).
> Block design was used, the task was repeated twice inside each session,
> with a baseline condition before, between and after the active epochs
> (Session 1: AA, Session 2: BB).
>
> We are interested in the comparison between the two conditions (i.e.
> between the sessions) A > B or B > A. So, our question is how to
> remove the session effects (i.e. the fact that the two conditions in
> the comparison are in two different sessions). We presume that it
> should be done by modeling the session effect as a user-specified
> regressor, but we are not familiar at all with this procedure.
This should be no problem because what you want is the session by
condition interaction. The session effects are simply the differences
in means between the two sessions, whereas you are interested in the
difference in activations. Simply proceed by modeling the activations,
as usual, in each esssion and compare the activations [with -1 1 1 -1
if rest periods are modeled explcitly (R A R B) or 1 -1 if not (A B)].
The session effect regressors are created automatically when you specify
the number of sessions.
I hope this helps - Karl
NB Session effects above refer to the main effect of session. The term
has also been used in reference to session x condition interactions
which, in this instance, is your treatment effect.
|