Hi,
Sorry, yes TFCE and F-tests have not yet been combined in randomise,
apologies.
Were the cluster-based F-test thresholdings different in extent or
just were some missing from the maxc output? If the latter it may just
be that they had P=1 so didn't show up? If that doesn't solve it then
maybe send the images to Matthew to debug?
Cheers, Steve.
On 16 Jul 2008, at 02:52, Kenneth Qiu wrote:
> Dear All,
>
> I'm using fsl-4.0.4-centos5_64 on Ubuntu 7.10 (64 bits) - the Gutsy
> Gibbon installed on a machine with intel core 2 Duo.
>
> I am trying to study DTI -reading ability correlation with TBSS.
> Basically, I have three groups of subjects with different age and
> the age spans within each groups are small. In order to find regions
> of the brain whose FA values are correlated with reading scores, I
> use the following design, where G1,2,3 represent the three groups
> and R1,2,3 are reading scores demeaned within each group.
>
> reading.mat
>
> G1 G2 G3 R1 R2 R3
>
> Sub1_1 1 0 0 a1 0 0
> Sub1_2 1 0 0 a2 0 0
> ...
> Sub2_1 0 1 0 0 b1 0
> Sub2_2 0 1 0 0 b2
> ...
> Sub3_1 0 0 1 0 0 c1
> Sub3_2 0 0 1 0 0 c2
> ....
>
> reading.con:
>
> /ContrastName1 EReadGrp1
> /ContrastName2 EReadGrp2
> /ContrastName3 EReadGrp3
> /NumWaves 6
> /NumContrasts 3
> /PPheights 4.500000e+01 1.130000e+02 5.400000e+01
> /RequiredEffect 3.955 2.500 3.995
>
> /Matrix
> 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
> 0.000000e+00
> 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
> 0.000000e+00
> 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> 1.000000e+00
>
>
> reading.fts:
> /NumWaves 3
> /NumContrasts 1
>
> /Matrix
> 1 1 1
>
>
> Since it is possible that the same brain region may having different
> correlate with reading among individual group, I decided to use F
> test. As Steve and Tom suggest TFCE is sensitive but slower, I
> decided to give it a try with small #perm. I used randomise -i fa_4d
> -m mean_FA_skeleton_mask -o reading -d readign.mat -t reading.con -f
> reading.fts -T -n 50 -V . I didn't see any tfce for F test but only
> for the three t contrasts. Is TFCE not yet implemented for F test?
>
> I then turned to cluster-based inference using randomise -i fa_4d -m
> mean_FA_skeleton_mask -o reading -d readign.mat -t reading.con -f
> reading.fts -F 12 -n 50 -V. I chose 12 as threshold based on my
> observation of a previous run of randomise . To my surprise, the
> cluster from reading_maxc_fstats did not match what I got by
> thresholding reading_fstats manually in fslview. And I think the
> cluster extents should be identical according to my understand of
> clustered based inference. Could anyone please help? Thank you very
> much!
>
> Sorry for the lengthy email, I try to give enough information for
> people to diagnose.
>
> Best,
> Deqiang
> University of Hong Kong
>
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|