Dear all,
I had a problem when I was trying to apply Christian Gaser's TFCE toolbox.
Details about the problem are below.
** What I have done
1. I ran a 2nd-level test on results of a whole-brain searchlight MVPA
analysis using the main SPM GUI: 'Specify 2nd-level' and 'Estimate'
2. Then I estimated the TCFE using the TCFE toolbox: 'Estimate TFCE' on the
TFCE GUI (5000 permutations)
- I selected the SPM.mat which was generated in step 1.
3. Then I applied the TCFE results to my data: 'TFCE results' on the TFCE
GUI
- I selected the type of statistic as 'TFCE', not 'T'
- I selected p-value adjustment to control for 'FWE' and p<0.05
** What happened
The result figure of the TFCE estimation (i.e., Distribution of maximum
tfce) looks fine, with my maximum values are above the p<0.001 threshold.
When I apply this TCFE output to my results using the statistic 'T' &
threshold p<0.05 FWE & cluster size K>50, 4 clusters are found, which are
reasonable in location and size. So far so good, but here comes the problem.
When I use the statistic 'TFCE' & threshold p<0.05 FWE, all voxels of the
whole brain pass the threshold, with all voxels grouped as one cluster.
Do you have any ideas about where this problem comes from? Alternatively, if
you could let me know at which step I did something wrong, I would
appreciate it very much.
Thanks in advance,
Junsuk
p.s. Please have a look attached files for detail.
