Hi,
> From my understanding of Nichols et al NeuroImage 25 (2005) 653-660 my
> desire to identify the areas of significant association (in at least one
> of the cohorts) that are not significantly different from each other means
> that it is OK to use the minimum statistic global null conjunction
> approach as implemented in SPM2. Or have I missed the point and should be
> using minimum statistic Conjuction null approach which I believe is
> implemented in SPM5.
No, I think you might be thinking of the 'interaction test'
conjunction implementation of SPM96 - but the conjunction tests as
implemented in SPM99 onwards do not address whether the comparisons
are similar in size, only whether they both pass some threshold.
Best,
Matthew
|