 |
 |
Dear Bonnie, This 'gigantic cluster' result from TFCE simply indicates that you have a diffuse effect, perhaps extending over the whole WM tract. If you did a univariate test on the global mean FA (e.g, in R or SPSS), you would probably find a strong significant result. If you wish to further understand this effect in your data, you can take an approach in VBM & PET analyses, and include a global mean FA covariate. With such a covariate in the model, any effects will be interpreted as "group differences after discounting global FA effects". It basically treats the global changes as a nuisance, and then lets you detect regional changes (that cannot be explained by the global changes). If you find no significant effects (positive or negative) it tells you that that the global changes explain all of the group differences. If you find a effect it tells you that you've found a regional effect that cannot be explained by global FA. -Tom On Wed, Jun 27, 2012 at 2:02 PM, Bonnie Y K Lam <[log in to unmask]> wrote: > Hi FSL users, > > > I am getting clusters as large as 70,000 voxels after running randomise > using the TFCE option. Should I be worried getting such a big cluster > linking almost all major tracts of the brain? > > I am aware of the cluster-based thresholding but am not sure if that will > help getting more specific clusters and am also not exactly sure of what > threshold (zstat value) I should be putting in. > > My randomise command is as follow for your reference: > randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d > design.mat -t design.con -n 5000 -x --T2 -V > > Thanks a lot, > Bonnie > -- __________________________________________________________ Thomas Nichols, PhD Principal Research Fellow, Head of Neuroimaging Statistics Department of Statistics & Warwick Manufacturing Group University of Warwick, Coventry CV4 7AL, United Kingdom Web: http://go.warwick.ac.uk/tenichols Email: [log in to unmask] Phone, Stats: +44 24761 51086, WMG: +44 24761 50752 Fax: +44 24 7652 4532