Dear SPMers
A couple of days I posted a query about making contrasts for a 2 session design (baseline; post-drug).
I am reposting in case this has gone undetected. Thank you in advance,
Pablo
----------------------------------
I have a query about how to modify my contrasts for 1 session study design to a 2 session design (looking at effect of dug).
My contrasts and study design for 1 session:
fMRI task has 2 within-subject factors (A) Task, with two levels: emotional question 1 or cognitive question 2 and (B) Emotional valence, with three levels: (1) positive, (2) neutral and (3) negative. Matrix design is in order of valence by Task:
A1B1 A2B1 A1B2 A2B2 A1B3 A2B3
To test (1) overall effect I set [1 1 1 1 1 1 ] for each subject and entered the resulting con images for each subject into one-sample t-test for the second level. I specified ‘1’ for F-contrast.
Step (2) test main effect of task used a one-sample t-test and fed with con images [1 -1 1 -1 1 -1 ]. Set ‘1’ for F-contrast. Also set a t-contrast for looking at differences between tasks.
Step (3) test main effect of Valence used a two sample t-test and fed with con images from two contrast: valence 1 [.5 .5 -.5 -.5 0 0] valence 2 [0 0 .5 .5 -.5 -.5].
For looking at the difference of the three valences I set F-contrast [1 0; 0 1] and as I have two rows I set 2 t-contrasts.
To test for (4) the interaction between Task and Valence I use a two-sample t-test and fed with con images from valence1 by Task [.5 - .5 -.5 .5 0 0] and valence2 by Task [0 0 .5 -.5 -.5 .5]. I then use a [1 0; 0 1] F-contrast and set 2 t-contrasts. One t-test will tell if any of the valences (valence1) depend on the Task type and the other if any of the valences (valence2) depend on the Task type.
My study now has 2 sessions (baseline; post drug) Importantly I anticipate an effect of drug, and also anticipate this effect would be modulated by task and valence. So I wish to simply add drug in my design but keep everything as the previous design. In terms of preprocessing I have concatenated baseline and post drug scans by individual subject.
What I would like to ask: if the following design make sense. To address this I think I should use the contrasts of the 1 session design and mirrored but with opposite symbol.
Matrix is duplicated now: S1A1B1 S1A2B1 S1A1B2 S1A2B2 S1A1B3 S1A2B3 S2A1B1 S2A2B1 S2A1B2 S2A2B2 S2A1B3 S2A2B3
Test (1) Effect of drug effect set as [1 1 1 1 1 1 -1 -1 -1 -1 -1 -1]
Step (2) test main effect of task used a one-sample t-test and fed with con images [1 -1 1 -1 1 -1 -1 1 -1 1 -1 1].
Step (3) test valence effect [.5 .5 -.5 -.5 0 0 -.5 -.5 .5 .5 0 0] valence 2 [0 0 .5 .5 -.5 -.5 0 0 -.5 -.5 .5 .5]
To test for (4) the interaction between Task and Valence I use a two-sample t-test and fed with con images from valence1 by Task [.5 -.5 -.5 .5 0 0 -.5 .5 .5 -.5 0 0] and valence2 by Task [0 0 .5 -.5 -.5 .5 0 0 -.5 .5 .5 -.5].
If anyone would be able to confirm whether these changes make sense I would be greatly grateful
Many thanks Pablo
|