As an alternative to taking to the con images from a [1 -1 -1 1] contrast to a one-sample t-test, you could take your (A>B) and (C>D) con images to a paired t-test at the second level. This may seem more transparent to you.
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] On Behalf Of Soha Saleh [[log in to unmask]]
Sent: 08 February 2011 17:59
To: [log in to unmask]
Subject: [SPM] imcalc to do a contrast?
I have 4 conditions in my experiment (A, B, C, D).. I am interested in finding the neuro activity that is higher in A than in B, and also the areas where the difference in activity between A and B, are higher than between C and D. Simply it is (A>B) - (C>D). I did each of the contrasts (A>B) and (C>D) in SPM8 and then used imcalc to subtract them to get (A>B)-(C>D). I got huge activity and T value that is very high which made me suspicious about this procedure.
My questions are: Is this a valid procedure? Is there any precautions or limitations I need to be aware of?
Thank you, I appreciate any assistance or advice you give.