Hi - yes you really need to be using corrected p-values in order to be sure of significance.
tstats of 0.01-0.02 are VERY small, if you really mean that - I would not expect those to reach significance.

Also - An FA threshold of 0.1 seems rather low - this would normally mean that you are including a lot of quite variable more minor tracts, which will reduce the power of your stats.

Cheers.


On 10 Feb 2012, at 09:14, amad ali wrote:

Dear FSL experts,

I ran a tbss analysis between 12 vs 13 very selected patients (same disease but they only differ in clinical presentation).
My parameters were :  FA threshold = 0.1, TFCE option and 10 000 permutations.
I've got very intersting results (corresponding exactly to our hypothesis) but only in the raw t-statistic image (tstat1 < 0.01 and tstat2 < 0.05). Unfortunately, I've got no results in the tbss_tfce_corrp files.
Do you think that my results are usable like that or that they are too open to criticism ? 

Thank you very much in advance for your answers and your comments !

Best regards,

Ali Amad


---------------------------------------------------------------------------
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
---------------------------------------------------------------------------