I have a (hopefully) simple statistics question. I have a dataset of reconstructed microelectrode recording sites across patients within the subthalamic nucleus (STN) - for each recording/site, I have a statistic for two analyses (testA and testB). What I want to understand is whether testA or testB values vary by anatomical location within the STN (is testA higher in one part, and testB higher in another?). To do so, I have been running a paired 3dttest++ (H0: for each voxel, are the testA and testB statistics the same?).
3dttest++ -setA *testA* -setB *testB* -paired -mask stn_mer_mask.nii.gz -ETAC 8
…where stn_mer_mask.nii.gz is a mask for all voxels with microelectrode (mer) recordings.
While this is producing encouraging results, I want to be sure that this is indeed the appropriate test. My primary concern is that there is an unequal spatial distribution of recording sites within the mask (microelectrode recordings just happen, we can’t record from every voxel in each brain). I would imagine this is analogous to voxel-based lesion-symptom mapping, but I wanted to know if this is accounted for with ETAC/ClustSim.
Thank you for your help!
PS - I have similar results/problems with running 3dttest++ on electrocorticography/surface data, and I was wondering if scripts generated by slow_surf_clustsim.py would also address this sparse spatial density issue.