Hi Michal,
On Thu, 4 Mar 2004, Michal Kuniecki wrote:
> Yesterday I set up another six analyses to run over night. This time I
> included only one simple contrast per analysis (to avoid possible matrix
> deficiency) and I also set Thresholding to “NONE”. Again I have got no
> activations, not a single bright pixel in the output ! I suppose that at
> least the analysis taking into account half of the group [1 0] or [0 1]
> should yield some results since combined analysis gives very large
> activations.
If you select "none" for thresholding you will get no thresholding and no
overlays - ie no poststats in effect. What you want instead is to select
"uncorrected" and change p to 1.
> Well the problem is that I have rather large N, at least for standards in
> fMRI measurement. I have 44 subjects in total, half of them are females and
> half of them are therefore males. Therefore I do not really think that the
> problems I experience are related to small sample.
fair enough, though OLS doesn't use separate variance estimates for the
two groups, so you still might gain modelling power by using FLAME and
setting up the group memberships.
> I realize that including the full setups of the analyses I did was not very
> useful, therefore I include the brief information about my data below
>
> The analysis yielding large activations (mean activation for entire group):
>
> In EVs:
> Input: Group: EV1
> 1 1 1 (female)
> 2 1 1 (female)
> 3 1 1 (female)
> …..
> 42 1 1 (male)
> 43 1 1 (male)
> 44 1 1 (male)
>
> In Contrasts&F-tests:
> EV1 = 1
>
>
> The analysis yielding no activations:
>
> In EVs:
> Input: Group: EV1 EV2
> 1 1 1 0 (female)
> 2 1 1 0 (female)
> 3 1 1 0 (female)
> …..
> 42 2 0 1 (male)
> 43 2 0 1 (male)
> 44 2 0 1 (male)
>
> In Contrasts&F-tests
> EV1 = 1 EV2=-1 (females-males)
>
> OR
>
> EV1 = 0 EV2 = 1 (only males)
>
> OR
>
> EV1 = 0 EV2 = 1 (only females)
I presume you mean [1 0] for only females. But apart from this, yes this
all looks correct. You could check your unthresholded stats/zstat*.hdr
images to see if they are nearly similar to the 1-group model.
Cheers.
Stephen M. Smith DPhil
Associate Director, FMRIB and Analysis Research Coordinator
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|