Hi,
My event-related design is simple two-conditions functional localizer. I want to
get a listing of voxels, which were activated significantly different by two
conditions (this is part of larger algorithm, so I do it manually). I don’t make
any multiple-comparison corrections.
My steps are:
1. Extract all data points with spm_read_vols and normalize the data
2. Take my design file and shift it to accommodate for hemodynamic lag. Since
my TR=2, I drop first three volumes and assign first design condition to volume
number 4. Last three design trials I drop as well.
3. Run anova1 matlab function for each voxel and count the number of voxels,
which received p<.05.
My problem:
I am getting only few hundreds significant voxels, while when I run this design
in SPM5 I am getting few thousands. Scrambling design files gives the same
few hundreds, so they are really junk. Do you see any problem with my
methodology?
Thanks a lot!
Daniel
|