I noted this when I was trying to contrast events with differing
durations. One event was about 200 ms and the other about 1200 ms.
The expected BOLD response should not have differed between the two
events nearly as much as the model did when I looked at my design file.
I scaled the short amplitude response and found that the data fit my
model much better and that the models for 200 ms and 1200 ms duration
only differed by their lengths.
On Apr 27, 2004, at 11:41 PM, Joseph Devlin wrote:
>> My experience is that anything below a one second duration stim time
>> will need to be scaled up in amplitude by the impulse's ratio to a
>> minute.
>
> That's interesting. I've used event durations of a 100ms without
> modulating the amplitude very successfully in the past. I wonder if it
> depends on how much variability you have in event durations? In other
> words, if all the events are essentially quick (100ms) then the
> amplitude
> of the evoked HRF only varies within a relatively small range. If
> these
> events are mixed with blocks of say 30s, however, then the difference
> in
> amplitude would be enormous and the fast events may very well jut look
> like
> noise in the design matrix. In this case, though, you probably
> wouldn't
> want to rescale the amplitude of the events because the difference in
> amplitudes reflects the important difference in the expected strength
> of
> BOLD signal, with much longer events (ie blocks) leading to
> significantly
> larger signal (and hence greater sensitivity in the design).
>
> I hadn't specifically considered this point previously, though, so I
> wonder
> what others think?
>
> Joe
>
> --------------------
> Joseph T. Devlin, Ph. D.
> FMRIB Centre, Dept. of Clinical Neurology
> University of Oxford
> John Radcliffe Hospital
> Headley Way, Headington
> Oxford OX3 9DU
> Phone: 01865 222 738
> Email: [log in to unmask]
>
|