On Fri, Jun 28, 2013 at 2:24 PM, fMRI
<[log in to unmask]> wrote:
Hi Donald,
How about having a second and third order effects in the parametric modulations. My understanding of these is that they also show whether there is a non linear relation between the BOLD signal and the parametric regressors.
>> The relationship is always linear with the PM. However, you are right that you can model non-linear relationship if you square or cube or transform the PM values before entering them as the PM. You can think of the beta of the PM as the scaling of the PM term - whether is the slope, parabolic, cubic or other effect. Hope that clarifies the issue.
Regards,
AS
On Fri, Jun 28, 2013 at 11:48 AM, fMRI
<[log in to unmask]> wrote:
Dear Donald and all,
Would it be correct if I have two sessions with different (experimental designs and thus parametric effects) and compare contrast of parm1 > parm2?
>> If parm1 and parm2 come from different sessions, you have to be careful how you compare them as you might not find differences if the baseline shifts between runs.
The difference is only in how subjects perform a movement.
The other question is : if I have three conditions : mov hard , mid and low.
I am recording subject responses and there are lots of variabilities. I thought to follow some previous papers where they model it as a one condition and including a parametric modulation. Would this be better than specifying three conditions?
In PM, you assume a linear relationship with the PM. If you model it as 3 conditions, then you can test different relationship across the three conditions.
Regards,
AS