I have a question regarding the correct statistical technique to use in SPM
to analyze data associated with a parametric regressor.
First, here's a little bit about my task:
I have 4 conditions listed in the model of my task: Condition A, Condition
B, Control Condition, and Fixation.
I have already looked at t-tests comparing activation between sets of these
conditions (i.e. ConditionA-B, A-Control, and B-Control).
In addition, I am constructing another model in spm that has the previous 4
conditions and a parametric regressor for Condition A. This regressor is
the subject's response during this condition (ranges from 0 to 6). So this
is the same model as above except in addition Condition A gets one
parametric regressor (and included in this are both 1st and 2nd order poly).
Now basically I am interested in whether activation in Condition A is
related to the parametric regressor (both 1st and 2nd order terms).
I'm not sure what the best way to look at this would be.
I could do F test with the following: [1 0 0 ...], [0 1 0 ...], [0 0 1...]
where the terms are [Condition A, Con. A Parametric Regressor Poly 1, and
Con. A Parametric Regressor Poly 2...], but I'm not sure if this is correct.
But I have heard that I should always subtract another condition out when
doing f test, so that data will show activation associated with a condition.
Is this correct? And if so, how would I do this with F test and what would
be best to subtract out?
Another possibility is masking the F test with something else. Would this be
correct, and if so what would I mask it with?
Thanks. I'm a relative novice with this and could use some advice.