Dear Saaussan,
> In a block fmri study I have one condition (B) and a rest condition (A).
> I used linear parametric specification to model the condition B. If I
> understand this correctly, this means that I used two basis functions
> to model condition B (constant=zero order, and linear function=first
> order). I did not specify expilicitly the rest condition (just entered
> into spm99 that I have one condition trial 1, which is equivalent to
> condition B)
>
> My questions are ,
>
> How does SPM99 represent the rest condition??
It does not. There is no need to represent it explicitly. It is
represented implicitly as the difference between the fitted activation
effect and the constant term.
> Does it assume that the rest condition A is modeled by a two basis
> functions similar to condition B ??
No.
> Does it assume that the beta value corresponding to the first order basis
> of condition A is zero??
Yes (if that effect were included in the design matrix).
> More specifically? If I use a T-test what do the following contrasts mean
>
> 1 0 (does this contrast compare the zero order between B & A)
Yes, in a sense.
> 0 1 (does this contrast compare the first order between B & A)
No. It tests for the contribution of the second (parametrically
modulated) regressor associated with B.
> what part of spm99 takes care of modeling the rest condition if it is
> not specified explicitly?
The rest and active conditions engender a differential response. Only
this differential response needs to be modelled. It is sufficient to
model B to model this differential.
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|