Dear Jorge,

> Thanks a lot for your answer regarding errors in the predicted volume
> of an activated cluster. I have a further question regarding cluster
> comparison.
> We've used a working memory paradigm which gives strong activation in
> the posterior parietal cortex (PPC) and the dorsolateral prefrontal
> areas, bilaterally. We performed the study with two subject groups.
> What I would like to do is to compare the volume of the activated
> clusters in the same active regions (eg., PPC) but for the two subject
> groups.
> In some sense the clusters I want to compare are not so 'different'
> because they represent activation in the same brain area for the same
> task. Is this approach valid? You mentioned that comparing clusters is
> not that easy. Did you mean comparing clusters from different brain
> regions?  If so, and although I am not doing it now, how would you
> recommend doing such a comparison?

1) Generally, comparing the activation between two groups would proceed
by teating for the goup x condition interaction and then making
inferences based on the ensuing SPM.  Taking activation size as a
meaningful index of activation and entering this into a meta-analysis
(e.g. two-sample t-test) is possible but not ideal.  As Eric pointed
out the cluster size not a measure of area activated (to do this you
would have to define 'activation' and threshold the appropriate
contrast images.

2)  To compare different brain regions one would have to extract data
(or statistics) from SPM and enter them into a standard stats package.
Again you could do a meta-analysis but generally one would use the raw
time-series and look for region x condition interactions (by assuming
both regions had the same error variance).

I hope this helps - Karl