Dear all,
Suppose I have a parametric model like this: There are altogether 2 conditions, which have a common parametric modulator P, say RT. And I intend to use canonical HRF and its temporal derivative as basis functions. So the Design Matrix is sth like:
cond1*bf(1) cond1*bf(2) P^1*bf(1) P^1*bf(2) cond2*bf(1) cond2*bf(2) P^1*bf(1) P^1*bf(2) [For simplicity, assume for now that I am only interested in 1st-order effects]
Intuitively, when more than 1 basis function is used, one would an F-contrast to capture the activation modelled by canonical HRF and temporal derivative.
However, in the case of parametric modelling, one might not only be interested in the brain regions whose activation changes regularly as the parametric values change but also in how the activation changes, i.e. monotonely increasing or decreasing in linear case.
Therefore, the betas have to be tested more specifically using t-contrasts because to what I know, only t-statistics reveal the positive or negative nature of the betas.
So how do I solve this problem? Any comments or advice will be greatly appreciated, many thanks!
Ce
|