Dear Bernhard,
SPM constructs all of its stimulus functions, whatever the length, by
specifying a single transient/delta/'stick' function per scan. This ensures
that the same code can be used to create events (single stick functions) and
boxcars ( lots of stick functions , one per scan, continuous for time (t)
give you a boxcar of length 't'). The confusion arises when the terms that
the GUI uses are inconsistent. You need to pretend that your variable epoch
studies are, in fact, event-related, so that SPM will create your variable
boxcar lengths using a train of stick functions.
So, taking your example:
> we are trying to specify a bloc design fMRI analysis with variable
> durations within trials. When we enter variable durations:yes, SPM asks
> to specify the number of scans. Later on this number (vector) has to
> be specified again. The resulting design matrix seems to make no sense to
> us.
Exactly. I had similar problems with this myself. You need to enter:
variable durations:yes
number of scans: enter your vector of epoch lengths here.
then - when SPM asks you if this is a epoch or event related study, choose
event. If you specify 'hrf only' as your basis function, you will effectively
get your variable length boxcar convolved with the canonical hrf, which I
think is what you wanted.
Best,
Dave McG.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|