Dear Sherif,
>I am not sure how to reconcile, in my mind, the following comments that you
>wrote to Tatia:
>
>1) to identify regions of CONSISTENT activation in ALL THREE fMRI studies, a
>conjunction analysis was performed using an SPM of the minimum t-statistic
>over the individual contrasts from the three groups
>
>with
>
>2) A significant effect (p <0.05) suggests that ONE or more contrasts were
>significant.
>
>The second comment mentions that only one of the X number of contrasts can
>make a region significant in this kind of analysis while the first comment
>refers to regions that are consistently significant over all studies... As I
>am in almost exactly the same situation as Tatia, I would very much
>appreciate your input...
There is a subtle but important distinction between
"consistent activation" and "consistently significant activation"
A conjunction is like a one-tailed F-test. It tests for any treatment effect
in any [of several one-dimensional] contrasts, but is only interested in
effects
that are consistently positive over all contrasts. If significant, one can
conclude
that one or more effects are likely to be present. If all the effects are,
by design, the same that the probability of an effect being expressed in any
contrast is unlikely to be zero.
This is not the same as declaring the activation in all contrasts to be
jointly
significant. To infer this you would simply apply a corrected threshold to
each
component SPM. There has been a recent discussion about this that will appear
in two papers in NeuroImage next month.
I hope this helps - Karl
|