Dear Wenrong,
Basically, all regressors resulting from the Fourier (or gamma) modelling
are "equally important", as they all together are fitted to your data to
explain maximum variance with minimum error. The way to look at what
variance they all explain together, you have to use an F contrast (an
identity matrix spanning the number of regressors modelling each condition,
e.g. if you have 6 Fourier terms [regressors], then use an eye(6) in the
appropriate place). If you finally want to get an idea of the shape of your
actual [ly fitted] response, you have to use the plot option in SPM. Because
of the voxel-wise approach, the response can be different for different
regions.
This is unlike modelling the HRF and temporal derivative (TD), when in some
circumstances you may just test for the HRF with a T-test (and not the TD),
if you consider the TD as modelling temporal shifts as a confound and
consider the HRF, testing for the size of the effect, as your effect of
interest. Again in contrast to the above, the shape of the response will be
very similar in different regions because you asked SPM to look for an
HRF-shaped response in the first place (with the TD allowing some limited
variation).
Hope this helps,
Helmut
----- Original Message -----
From: "Choo Wen Rong" <[log in to unmask]>
Sent: Tuesday, February 06, 2007 5:19 PM
Subject: f or t constrast for basis functions such as foruier and gamma?
hi all SPM users all there,
for basis function like gamma and fourier functions used in the "specify
1st level", do i have to use a particular contrast like "f contrast"
during "result" part in spm 5 to analyse the data ?or that i can have to
use "t contrast"? any comment will be greatly appreciated.
wenrong
|