Dear SPMers:
We are examining a PET rCBF data set comprising two groups of patients,
under two conditions with two replicates of each condition. For examination
of main effect (and group effect) of condition and of the group by
condition interaction, we have instantiated a partitioned 'two study'
design in SPM96, employing subject specific ancova.
The first question is simply, are the cluster level significance values
(which we understand to derive from a conjoint assessment o f peak height
and spatial extent of the spm map) in any sense corrected for multiple
comparisons, or should they only be reported after thresholding for peak
height (p<0.001) and spatial extent (p<0.05)?
The second question relates to the partitioned design of the two study
matrix. If we wish to examine the 'main effect of group' i.e. between group
differences across both conditions or indeed under each condition
separately, it is my understanding that this can only be done under a
'degenerate design' i.e. single-subject, replication of conditions. This
reveals consistent findings of interest. The problem here is that in this
model we lose the facility to employ a subject-specific ancova, and the
pairwise nature of the effect of condition is also lost. Is there any way
round this problem?
Many thanks for your help,
Cindy Fu
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|