Dear Wil,
At 08:43 PM 4/4/01 +0100, you wrote:
>I am analyzing an event related design, and would like to model the time
>derivatives for the hrf. It seems the most appropriate way to do this is to
>use the F-Contrast.
Yes. When comparing conditions that are modeled with several functions,
F-contrasts are appropriate.
>For example (1 run, 2 event-type study)
>1 0 0 0 0 0
>0 1 0 0 0 0
>0 0 1 0 0 0
>0 0 0 -1 0 0
>0 0 0 0 -1 0
>0 0 0 0 0 -1
There are two things wrong here:
i) The design specifies three columns, when there should be two (hrf and
its temporal derivative)
ii) The structure of the F-contrast should enable one to compare between
corresponding columns in the same rows. The appropriate structure would
therefore be...
c1 c2 (conditions 1 and 2)
h t h t (h=hrf; t=temp.derivative)
1 0 -1 0
0 1 0 -1
>When I get my composites, it seems that I get a map with acivations
>representing any difference between the groups (t squared). How can one pull
>apart the (greater) activation for event-type 1 separately from event-type
>2?
The contrast I have suggested above will pull out voxels in which activity
in the two conditions is 'different', either with respect to time, or with
respect to amplitude, or with respect to some combination of the two. To
pull out voxels which only have a greater amplitude, one might specify the
following t-contrast...
c1 c2 (conditions 1 and 2)
h t h t (h=hrf; t=temp.derivative)
1 0 -1 0
However, in such a situation, one explicitely models the temporal
derivative as a confound. If you are only interested in amplitude
differences and not a combination of amplitude and time differences, why
not just model the hrf without the temporal derivative? This way, you will
also be able to specify the direction of the difference (see later points
below).
>Also, would one set up a contrast such as
>1 0 0 0 0 0
>0 1 0 0 0 0
>0 0 1 0 0 0
>to get the event related activity for the first event type?
More correctly, the F-contrast structure would be
c1 c2 (conditions 1 and 2)
h t h t (h=hrf; t=temp.derivative)
1 0 0 0
0 1 0 0
>And if so, is
>there a way to view the activations and deactivations separately?
F-contrasts will only select voxels in which there is a 'difference',
irrespective of the direction of that difference. However, it is still
possible to see the direction of the difference by plotting the fitted
responses for each condition in voxels revealed by the F-contrast. One of
the advantages that t-contrasts has over F-contrasts is that one is able to
specify the direction of the difference. t-contrasts would therefore allow
you to specify the direction of the differences, and to generate separate
SPM{t} maps for activations and deactivations.
Best wishes,
Narender.
