Hi Todd,
it depends on your model. However, two solutions might work: First, try melodic and look at the time course of the motor response. It's peak may give you an idea of the offset. Second, you could try FLOBs and reconstruct the fitted response using the COPEs and associated basis functions.
Hope that helps,
cheers-
Andreas
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Todd Thompson [[log in to unmask]]
Gesendet: Sonntag, 10. Mai 2009 21:49
An: [log in to unmask]
Betreff: [FSL] Unusual/backwards analysis help?
Hi, all. I'm currently analyzing data from an modified Erikson Flanker
task (fast event-related, one trial every TR), and I'm having a bit of
trouble because of what I suspect was an error in the data
acquisition. Most of my individual subject results look fairly good,
but there are some subjects for whom the activation maps look like
pure noise. I think what may have happened is that the stimulus
presentation was not correctly synchronized with the scanner, and my
model may be anywhere from 1s early to 6s late compared to the data.
However, I don't have a good way of demonstrating this, and it's poor
form to exclude subjects just because they look funny.
Here's my thought: in this task, responses were made either with the
right or left hand. If, in the "good" subjects, I run a model using
just the left-response and right-response as conditions, the
contralateral motor cortex is hugely active. In the "bad" subjects, I
again see noise. Is there a convenient way to fit the left/right
modelled regressors to the bad subject's data in a sliding window, to
determine where the best fit is between the regressor and the data? If
it turns out, for example, that if I run the left/right analysis with
a 1.5s offset the data looks perfect, then I'd have a good way to
salvage my original analyses, too. Or, at the very least, I'd have a
valid reason for subject exclusion. I'm just not sure of the best way
to figure out that offset.
Thanks for any insight!
Todd
|