Hi,
On 18 Feb 2009, at 08:27, Yuzheng HU wrote:
> Dear FSLers
> I intend to do unpaired two-sample t-test with IQ scores as confound
> regressor. I designed events and contrasts like this:
> ------------------------------------------------
> EVs
> Number of main EVs 3
> Number of addition, voxel-dependent EVs 0
> Group EV1(group_A) EV2(group_B) EV3(demeaned_IQ)
> 1 1 0 -9.7
> 1 1 0 1.2
> : : : :
> 2 0 1 12.2
> 2 0 1 -10.0
>
> Contrasts
> Contrast 2 F-tests 0
> for_group_A>group_B 1 -1 0
> for_group_B>group_A -1 1 0
>
> When I saved the design, there was a warning window jumped out with
> the
> following message:
> Problem with processing the model: Warning – design matrix uses
> different
> group (for different variances), but these do not contain
> “separable” EVs for
> the different groups ( it is necessary, that, for each EV, only one
> of the
> groups has non-zero values).
>
> My questions are,
> 1) Shall I treat all subjects as group “1”, or shall I split IQ
> scores as two
> events? If I split IQ scores as two events, shall I demean IQ scores
> separately in each group or across all subjects?
I'm guessing that you're using randomise, which doesn't actually use
the group membership information anyway (FLAME does in FEAT, that's
why this option exists in the GLM GUI). So yes just set all subjects
to have the same group membership.
> 2) In randomise command, I needn’t use option –D since I have demeaned
> IQ scores, am I right?
More like: because you are modelling the data mean with EVs 1 and 2,
you shouldn't use the -D flag to demean the data.
> There is a file named design.grp produced by Glm_UI,
> is –e option necessary in randomise?
Not in this case, no.
>
> Must I use –x option in randomise? May I use --T2?
-x gives you voxelwise inference but you probably don't want that -
you probably just want TFCE, using the --T2 option.
Cheers.
>
>
>
> Thanks a lot.
>
> sincerely, Yuzheng
>
> ========================
> Biox-lab, Zhejiang Univ. China
> ========================
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|