> Just a brief comment on contrasts for flexible basis set. I guess you
> could use a t-contrast if you are looking for an HRF of a shape a bit
> different than the canonical HRF (e.g. a bit earlier or broader). I
> sometimes find the F-contrast approach to be too flexible in that there
> is no way to make sure that the activation image isn't dominated by e.g.
> the second derivative of the HRF, and this is seldom what you want. I
> would be nice to be able to put some constrains on all the potential
> shapes provided by a basis set of e.g. HRF and its first and second
> order derivative. Does anybody have any ideas on how this could be done
> in SPM?
I think that if the HRF in your data differs so much from the canonical HRF that you have to use all kinds of dirty tricks, you could better use a different HRF. For example, a group mean HRF extracted from the region where you find activation (with an F test using the HRF and a few derivatives).
Alle Meije
|