Hi all,
I'm trying to pin down the interpretation for conjunctions, especially
SPM96 vs SPM99 conjunctions.
In particular, I realized that SPM99 conjunctions only address
consistency of rejecting the null, not consistency of the magnitude of
the response.
For example, say a SPM99 conjunction of two contrasts has a corrected
threshold of 3.01. Then a voxel will be significant if one contrast
has a t-stat of 3.02 and another contrast has a t of 10.0. But the
difference of these contrasts is likely to be significantly large.
Hence both contrasts are consistent in rejecting the null, but one
contrast has a significantly greater magnitude than the other.
In summary...
Approach: Single contrast equal to sum of intrasubject contrasts.
Answers
Question: Did the subject average 'activate'? Was the mean of
subject responses significantly different from zero?
Problem: Single subject can drive activation.
Approach: SPM96 Conjunction
Answers
Question: Did all subjects activate consistently? Was the mean of
subject responses significantly different from zero AND
was there *no* interaction between the intrasubject
contrasts?
Problem: Interaction threshold arbitrary; difficulties in
determining p-values.
Approach: SPM99 Conjunction
Answers
Question: Did all subjects activate? Are all subjects responses
significantly different from zero?
Problem: Subjects can be significantly different in magnitude.
What do folks think?
-Tom
-- Thomas Nichols -------------------- Department of Statistics
http://www.stat.cmu.edu/~nicholst Carnegie Mellon University
[log in to unmask] 5000 Forbes Avenue
-------------------------------------- Pittsburgh, PA 15213
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|