On Wed, 16 Sep 2009 16:57:11 +0200, Julia Weiler <[log in to unmask]>
wrote:
>Dear SPMers,
>
>I have a basic question on whole-brain interaction analysis.
>
>My task consists of two conditions (Cond1, Cond2) which both contain two
>phases (PhaseA, PhaseB).
I'll call the cells A1, B1, A2, B2, where e.g. A2 means condition 2, PhaseA.
>I computed F-contrasts with contrast weights=[1 -1 -1 1] in order to
I'll assume you ordered the columns as A1, B1, A2, B2. The interaction is then
A1 - B1 -A2 + B2
which we can rewrite as
(A1 - A2) - (B1 - B2)
>check for interactions between
>Condition and Phase and obtained several significantly activated clusters.
>I also looked at t-contrasts within each Phase, e.g., I computed
>t-contrasts between Cond1 and
>Cond2 for PhaseA and t-contrasts between Cond1 and Cond2 for PhaseB.
I'll write those as (A1 - A2) and (B1 - B2)
>My question is the following: The interaction reveals a cluster that
>shows stronger activity for Cond1
>in PhaseA, but stronger activity for Cond2 in PhaseB.
Sounds like you found a cluster for which
(A1 - A2) - (B1 - B2)
was significantly positive. In the above line, you're concluding that
(A1 - A2)
is significantly positive, and
(B1 - B2)
is significantly negative. But you can't conclude that. For example, it could
be that
(A1 - A2)
is "'quite" positive, yet not significantly positive, and
(B1 - B2)
is "quite" negative, yet not sign negative, yet their difference is significantly
positive.
>Should I then see
>the same brain region in the
>t-contrasts that I computed separately for the two phases?
No (see above).
>I.e. should
>the t-contrast for PhaseA
>reveal stronger activity for Cond1 (compared to Cond2) in the brain
>region that also emerged in
>the interaction analysis? And should the same brain region appear in the
>other t-contrast for PhaseB
>with stronger activity for Cond2 (compared to Cond1)?
>In my data, I obtain a number of clusters in the interaction analysis,
>which do not appear in the
>separate t-contrasts and I am wondering why this is the case.
>
>Any help would be appreciated - thanks!
>
>Julia
>
>
>--
>Julia Weiler
>
>Ruhr-University Bochum
>Inst. of Cognitive Neuroscience
>Department of Neuropsychology
>GAFO 05/606
>Phone: +49-234-3223574
>Fax: +49-234-3214622
|