On Thu, Feb 2, 2012 at 8:10 PM, Ben Becker
<[log in to unmask]> wrote:
Dear SPMers,
we used a blocked design fmri-paradigm to investigate intervention effects in a single-subject. The patient was scanned twice (t1 & t2); the design incorporated 3 social (s1, s2, s3) and a non-social control condition (c). To investigate intervention effects regarding the specific effects on social stimuli I’d like to use the following design:
1) First level design with two sessions & the following conditions:
s1(t1) s2(t1) s3(t1) c(t1) s1(t2) s2(t2) s3(t2) c(t2)
Would the following t-contrast reveal the stronger activity at t1 specific to social stimuli:
1/3 1/3 1/3 -1 -1/3 -1/3 -1/3 1 ?
No. The contrast tests the difference of social-non-social between t1 and t2. The test says nothing about the difference only being in the social conditions. Secondly, the word stronger coud be misleading. If social-nonsocial at t1 is zero and social-nonsocial is negative at time 2, then you'd get a positive effect. I try to use the word higher or larger.
2) If I would test a control group twice with the same paradigm: could I use the above mentioned interaction contrast in a two-sample
t- test (according to rik henson’s comparing a single subject against a group of controls)?
If you have the contrast above in a bunch of controls and then you have the same contrast for a single patient, then yes you could use his approach. Keep in mind that the results would represent a three-way interaction of stimuli*time*group.
Thanks in advance & best regards
Ben