I am conducting a PPI analysis within FSL comparing three groups. My problem is that when I went to go look at my within-group results, I discovered that the three groups appear to have almost identical results. The results are not 100% identical, but approximately 98% of the voxels are shared between the groups- way too many not to be an error.
Has anyone else encountered this problem or have any suggestions? I suspect that there is a duplication in the data such that many of the subjects from one group have been erronously copied to another group. The part that is confusing me is that for the results to appear this similar, I would think the duplication for the PPI analysis would need to be at the level of the whole brain (residualized data) and the extracted ROI time course. However, I don't see any duplications at either level.
Below are the steps I am using to conduct the PPI and possible areas that I have checked for errors
* Step 0- Run the original FEAT to generate example_func.nii data file
* Step 1- create a nuisance version of my main FEAT (step 0) in which the CSF and WM have been regressed out
* Step 2- add 10,000 to the resid4d whole brain data to generate resid4d_10000
* Step 3- extract ROI time series
* Step 4- Run the Individual subject level PPI- only statistical modeling (no registration)
* Step 5- Copy the registration folders (from Step1) into new PPI directories generated in previou step (Step 4)
* Step 6- Group level analysis- Inputs are individual level subject PPI feats from Step 4
Areas I have checked
* Made sure my input to the Group level anlayis (Step 6) were not incorrect
* Verified that the Step 4 - residualized4d_10000 files were not duplicates
* Verified that the ROI time series from Step 3 were not duplicated
|