 |
 |
Hi Steve, Thanks for your useful suggestion. However, I have another question about the multiple correction on TBSS results. I use the TFCE or -c 2.3 (or my method mentioned previously) to do, but I don't find significant location. But using the the cluster-based method in AFNI tools (Alphasim command), some location could be detected with the application of the cluster size corresponding to the alpha level (e.g. q < 0.05). Note that this correct procedure assume data have random distribution. I have two questions as follow. 1) I don't konw whether the FA values in the main skeleton meet the distribution. 2) if no, could I only use the cluster size (maybe differ from that provided by AFNI) to threshold the TBSS results. In the paper, we would indicate that this results are indeed corrected for multiple comparison. Many thanks. 2008/11/19 Steve Smith <[log in to unmask]> > Hi, > > On 19 Nov 2008, at 05:26, liang wang wrote: > > Hi FSLers >> >> If I would investigate the relationship between FA values in the skeleton >> voxels and multiple behavioral data together, I need to manually generate >> the design.mat and design.com files. However, I don't know how to set the >> parameters: PPheights in the two files and what roles the parameters in >> randomize process. >> > > Don't worry about these - you can just set to 1 or leave this line out of > the header - randomise doesn't use it. > > Also, if I want to select the cluster-based thresholding corrected for >> multiple comparisons, how to set the required t (or F) value. At present, >> according to the no corrected results (e.g. file_vox_corrp_tstat.nii.gz >> obtained from 5000 randomize times),I select a threshold corresponding to >> the considered significant level (e.g. p < 0.001). Thus, using the threshold >> for t or F contrast (a parameter of -c or -C), I rerun the whole analysis >> procedure. I don't make sure whether the selection step is suitable. >> > > Indeed - that's the problem with cluster-based thresholding - there's no > principled way of choosing the initial cluster-forming threshold. You might > like to try some arbitrary threshold like 2.3 (or 0.000023 or 23000), which > seems popular; alternatively you can try using TFCE instead, which gets > around this problem, by setting the -T option instead of -c. > > Cheers. > > > I am appreciated for any suggestion. >> >> Best >> Liang Wang >> > > > --------------------------------------------------------------------------- > Stephen M. Smith, Professor of Biomedical Engineering > Associate Director, Oxford University FMRIB Centre > > FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK > +44 (0) 1865 222726 (fax 222717) > [log in to unmask] http://www.fmrib.ox.ac.uk/~steve<http://www.fmrib.ox.ac.uk/%7Esteve> > --------------------------------------------------------------------------- > -- Liang Wang, Postdoc Woodward Lab Department of Psychiatry University of British Columbia BC Mental Health & Addiction Services 938 West 28th Avenue Vancouver BC V5Z 4H4 Telephone: 1-604-875-2000 (ext. 4735) Fax: 1-604-875-3871 Email: [log in to unmask]