Re: [Marsbar-users] ROI extraction with time derivatives

[Carolyn L. F. <cl...@ma...>]

Thank you, Matthew.

This is an event-related design, though, not block, so is it still appropriate to ignore the time derivatives and just do A-B?

-Carolyn


Carolyn L. Fort
Research Specialist, Sr.
Department of Psychiatry
University of Arizona
PO Box 245002
Tucson, AZ 85724-5002
v: 520-626-8568
f: 520-626-6050

 
On Monday, June 25, 2007, at 03:56AM, "Matthew Brett" <mat...@gm...> wrote:
>Hi,
>
>> I'm extracting ROI timeseries data from an old dataset which uses time derivatives in the event-related model. As a result, when I extract the ROI data I have 2 betas for each condition. I'm not sure how to make sense of these data. For example, in a box-car model with a simple ABABAB design I would have 6 datapoints and would subtract the B's from A's for the contrast A-B. For this dataset I have, for example, AaBbAaBbAaBb (lowercase being the time derivative output). How do I calculate A-B here?
>
>The simplest way would be to ignore the time derivatives, and just do
>A-B for the time derivative model.  That might not be very good
>though, because in block designs the time derivative is correlated
>with the HRF, and you might well lose power.  So, it might be best to
>use the marsbar interface to get a % signal change for your block and
>subtract %s for A and B,
>
>Best,
>
>Matthew
>
>