Cyril,
> The design matrix is a 6 conditions within subjects ANOVA with 12
> subjects. I performed several contrasts like 1>4, 2>5 and 3>6. Next,
> I performed the conjunction (1>4)+(2>5) that gave me a 1st result. In
> the same time, SPM created a new contrast that corresponds to this
> conjunction.
>
> When I look again at the conjunction using this contrast, the result
> is different than the one obtained by (1>4)+(2>5). I tested with other
> conjunctions .. same results.
SPM only creates a new contrast if you're assessing the global null
and the contrasts are not orthogonal (Global null conjuntions require
independent test statistics).
The two contrasts for (1>4) and (2>5)
[1 0 0 -1 0 0]
[0 1 0 0 -1 0]
are orthogonal, and your ANOVA design is also orthogonal, *except*
that the nonsphericity whitening has yeilded the design matrix that
isn't perfectly orthogonal. (You are accounting for nonsphericity, yes?)
So I belive that the explanation is that nonsphericity has 'colored'
your ANOVA design, and that makes the design space of your contrasts
non-orthogonal, forcing the creation of new a contrast when you request
Global Null conjunction inference.
One 'solution' (if this is a problem), is to use the Conjuction Null,
which doesn't require independence of test statistics.
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
|