Hi - you should subtract the means of EVs 3 and 4 so that they are both zero mean.   There is also no need to set the "group" memberships for this model, as randomise doesn't do group-wise variance estimation, and it can work out the correct permutations from EVs 1 and 2.

Cheers, Steve.


On 31 Oct 2009, at 17:33, 김지현 wrote:

Dear Prof. Smith and other experts
 
Thanks for the replies and... sorry for bothering you again.
As for the statistical model of two sample unpaired t-test with 2 nuisance variables (age, sex) in TBSS FA analysis...
By using GLM GUI in FSL 4.1.4, I have set up the following design matrix (pls see the attached figures).
And the 'design.con' file includes;
/ContrastName1 control > patient
/ContrastName2 patient > control
/NumWaves 4
/NumContrasts 2
/PPheights  1.067873e+00 1.067873e+00
/RequiredEffect  1.617 1.617
/Matrix
1.000000e+00 -1.000000e+00 0.000000e+00 0.000000e+00
-1.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
 
I want to be sure that I am right, or please let me know the correct way.
Thank you in advance for your kindness.
 
Kim
 
<design2.png><design.png><contrast.png>


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