Hi all

Here is my story:
I have 2 groups of subjects.
I transformed each individual subject to Talairach and processed each of them with fico.
I took the Fisher transformed of each fico map using 3dcalc
(3dcalc -a 'group1_subject1 _cor+tlrc[3]' -expr 'atanh(a) ' -fscale -prefix group1_subject1).
I used 3dttest to run group analyses, first on each group separately, then comparing both groups.
I also performed a ROI analysis, extracted the signal in selected ROIs (before fico but after Talairach transformation) from each individual from the two groups and calculated correlations with Excel.

The group maps on each group made total sense to me. Group 1 exhibited mostly positive T values in agreement with the ROI analysis. Group 2 exhibited both positive and negative T values in different regions, also in agreement with the ROI analysis. The ROI analysis also showed that in regions where both groups had positive correlations, those correlations were always stronger for group 1 than for group 2.

However the results of the analysis comparing both groups puzzled me.
Here is the basic command I used:
3dttest -prefix ttest_atanh_gr1-grp2 -1blur_fwhm 6\
-unpooled -set1 group1subject1+tlrc ….
-set2 group2_subject1+tlrc ….

This group analysis produced a map with almost only negative T values, which I interpreted as group 2 being more strongly activated in those areas than group 1 (right?). However, this really doesn’t fit with either the group maps acquired separately on each group nor with the ROI analysis.

Is there some statistical trick that I don’t understand, or does that result make sense? or did I just do a very stupid mistake in my commands or worse in my reasoning?
Any help will be strongly appreciated.

Thank you very much

Barbara