Hi,
On Thu, Aug 4, 2011 at 4:37 PM, David ro <[log in to unmask]> wrote:
> Hello SPM'ers!
> Since nobody answered yet the question that I sent yesterday, I thought that
> maybe a clarification would be helpful here, and hopefully someone will pick
> it up...
> The problem is as follow:
> The default number of bins (for TR =2) in Marsbar is 13. that could be
> changed manually, but I assume that this number of bins is given for a
> certain reason - possibly it is related to the HRF, which take a fixed
> amount of time since the starting point of the neural response.
Right, that's guessing that you want your estimate to last 26 seconds.
But obviously you can change the default if you want to assume longer
or shorter HRFs.
> If that's indeed the case, then it implies that what happens in the
> beginning of the trial/block is the only thing that matters to the time
> courses that we see. For example, if I had a trial that took 20 sec and I
> look for time courses for the whole time period, everything that happened
> after the first 1-2 sec (or more? or less?) does not really show up in the
> time courses. So question #1 is: Is that correct?
I don't think I understand what you mean. By default you'll only be
estimating 13 time points at 2 second intervals, so anything outside
that 13 * 2 = 26 seconds will not affect the time course for that
event, and anything within that time, will.
> In contrast, if the time course analysis does show the whole trial, then it
> should matter how many bins I put. So, for different lengths of trials, I
> should request for different number of bins. So question #2 is: Is that
> correct? and if so, why do we have the default of 13 bins?
Default explained above. Yes, if you think your event time course
lasted 40 seconds, you'd want 20 bins.
> If the latter argument (time matters) is correct, then here's some strange
> inconsistency that I noticed. Imagine that I ask for time courses for an
> event that took 20 sec. Since the default number of bins is 13, then
> anything that happened after 20 sec and appears in the time courses report,
> actually belongs to the next event in the experiment. But when I ask for
> time courses for the next event, the the percent signal change for the first
> sec does not match the percent signal change of the 21th sec of the previous
> event. So question #3: Why is that?
I'm not sure what you've done here. Have you modeled both event 1 and
event 2 in the same model? In that case, if your events start at the
beginning of the TR, the first few bins for event 2 will be identical
to the the last few bins for event 1 and the model can't sensibly
distinguish the signal due to each. The models for these bins are
collinear. The difference in parameters between them is only an
artefact of the fitting procedure.
> I hope that this clarification helps a little, and looking forwards to hear
> from the greatest minds that use Marsbar to shed light on this burning
> issues...
You should post to the marsbar list - as it is mainly me answering, it
can get slow when I'm busy, as I am now, but I will do my best,
Cheers,
Matthew
|