Hi AFNI users,

I would like to reproduce results published in Cole et al., NeuroImage 2010, in which the authors tried to identify the brain's most globally connected voxels.

The basic idea is the following: for each voxel, compute the wGBC score, which is the average of correlations of the given voxel with all other gray matter voxels. Identify as the most connected voxels those with the highest wGBC scores.

The main contribution of the paper is that it also uses statistical inference to ensure reliability across subjects. The authors employ a group ANOVA which compares each voxel's wGBC to zero and set the ANOVA threshold so that only the top 10% voxels remain.

I would like to implement a similar approach in AFNI. However, I feel that an easier approach would be the following:
1. For each subject, compute their wGBC map and convert it to z-scores (subtract mean, divide by standard deviation).
2. Run a one-sample t-test (using 3dttest++) comparing each voxel's wGBC score to zero.
3. Correct for multiple comparisons using 3dFDR.
4. Set a threshold so that only the top 10% voxels remain.

Does this approach seem sound to you?

Thanks in advance,
John



Edited 2 time(s). Last edit at 11/25/2014 09:17AM by afniuser.