Dear Alex, Em, and all,
> This may be similar to a problem we encountered when trying to analyse
> sparse data. It boils down to not being able to put the stimulus starting
> before the first scan (requires a negative onset). Matthew Brett may be
> able to comment further.
Nothing significant I can think of to add to Karl's response. Essentially it
seems to me you have a PET-like design (no significant temporal autocorrelation
expected, no low pass filter required), but you may well want to add a high
pass filter, which is not available in the PET design interface. Of course
you'll have to adjust the width of the filter to the TR the interface knows
about - in Karl's suggestion: 3.
Luckily I think Karl's suggestion gets around the problem that Emma is
referring to, which is that, in the FMRI design interface, you cannot specify
an event that started before the first scan. The reason this doesn't matter
is that you don't need to (should not) convolve with the hrf, so you can model
the events as occurring instantly.
Cheers,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|