Dear all
Though I have been searching the archive up and down I am still not
100% clear about the following question.
I have performed an efMRI study (2x2 factorial, but this is not so relevant
here) where we had to present instruction screens of about 10 s duration.
I see three ways to deal with this fact
1) model the instruction as a condition, effectively reducing my dfs, but
also increasing the model fit (at least I hope so)
2) model the instruction as a regressor of no interest
3) ignore the instruction in my model
Apart from the fact that I am not sure which way is the best (I guess this
also has should be determined empirically, no?), I have two questions:
1) how would I enter my regressor of no interest? SPM asks for
a 100 scan vector (1 session=100 scan). Would this be something like
1 1 1 1 1 0 0 0 0 ... 0, assuming that the first 5 scans represent the
instruction? thus 1 specifiyng regressor is 'on', 0=off
2) If this approach is correct: how would the model deal with the fact then
that due to the sluggishness of the hrf I do not only have influences of the
instruction screen on my data during the time the instruction screen is one,
but also later on; wouldn't this fact speak for modeling the instruction
as a condition since then it would be convolved with the hrf?
any help and comments are gratefully appreciated
claus
|