1. How do we set up a conjunction analysis where we ask whether a particular voxel is signficant across both contrasts (ie, hits-misses for the jittered group and hits-misses for the fixed ISI group)?
If you did the hits-misses contrast for each subject in 3dDeconvolve, you might run 3dttest (-unpaired if the subjects in the two groups are different) or 3dANOVA2 to compare the two groups.
2. If we were to set up a more traditional ANOVA, would we use 3dANOVA where the input -a is jittered duration v. fixed duration and subjects are assumed to be part of a fixed-effects model OR would be use 3dANOVA3 type 5 where the input -a is as above, and -b is hits-baseline or misses-baseline, and -c is subject. We would then ask for the -bdiff of factor -b.
If you want to run ANOVA with only interest in the contrast of hits vs misses between the two group, it would be 3dANOVA2 (one factor is group, and the other either the contrast). However, if you want to check whether there is any random effect for individual subject, run 3dANOVA3 -type 5 with the 3rd factor being the subject.
3. With regard to question #2, we have assumed that it is more desirable to carry our analysis as far as possible in individual subjects, and therefore use the glt combination of hits-misses for each subject as the input into our ANOVA, rather than hits-baseline and misses-baseline. We don't think that approach would work with 3dANOVA3.
I'm not so sure what you exactly mean here. You can run 3dANOVA3 -type 5 with the 2nd factor being hits-misses, or hits, or misses separately, but you can't do it with hits and misses together as the 2nd factor if this is what you refer to.
Gang
