If you can legitimately justify it, why not use 240 for all? Using 120 for some subjects might cut out task-related signal.
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] On Behalf Of Jacques Temeran [[log in to unmask]]
Sent: Thursday, November 04, 2010 4:40 PM
To: [log in to unmask]
Subject: [SPM] Pseudo-random design and high-pass filter issue
Hi to all,
I have a pseudo-random box-car design (order of conditions varies between subjects).
A typical run goes like this (epoch duration = 30 sec, TR = 3 sec):
A r A r B r B r A r B
Where:
A = Condition 1
B = Condition 2
r = rest/baseline
I modelled r implicitly and I am only interested in the A minus B contrast.
The problem is that my estimated high-pass filter cutoff is 120sec or 240sec, depending on the order of conditions within a run (as these vary per subject). Should I use different values (i.e. depending on the subject) or settle for one value for all?
Note that most of my subjects end up with an estimated 240sec cutoff (from formula as well as from looking at the frequency domains) and that using this 240sec cutoff for all leads to more significant (and closer to the literature) results than using a 120sec for all.
Any thoughts are welcome,
Jacques
|