Hi,

We have a question about using a common baseline in conjunction analyses.

Our fMRI study involved 2 closely matched conditions that were compared to a single common baseline. There has been talk, particularly with SPM users, that there is a need for conjunctions to have independent (or 'nearly' independent) baseline tasks. It was suggested to us that we should randomly divide our baseline and compare each task to half the baseline for the conjunction analysis.

We found a posting on the SPM message board to this effect from Karl Friston: "Conjunctions are only really useful when the two contrasts are [nearly] orthogonal. Because you have a common baseline you cannot use a conjunction analysis unless you split your baselines into two giving four conditions A1 A2 B C. You could then ask where B > A1 and C > A2
[(-1 0 1 0] and [ 0 -1 0 1])."

In surveying the literature, it seems that few SPM users apply this principle and even fewer AFNI users.

We are not sure whether SPM and AFNI use the same data for determining the overall variance particularly in the fixed effects stage. Moreover, is it true that AFNI, but not SPM, considers all of the volumes in a data series whether they participate in a particular contrast or not when fixed effects are calculated?

We would greatly appreciate some direction on this matter.

Thanks very much,
Jenny