Hi Isabell,
yes, your final level design is of full rank so does not generate any
residuals needed for testing significance. The easiest thing to do is
to run a simple two level GLM, where at the first level you have your
single sessions and at the second level you use
time1 run1
time1 run2
time2 run1
time2 run2
time3 run1
time3 run2
as six input files and use a
1 1 0 0
1 1 0 0
1 0 1 0
1 0 1 0
1 0 0 1
1 0 0 1
design matrix with your old contrasts. Note that the first column is
all ones, i.e. here you assume that the residual variance before/
during and after treatment stays the same. If you were to have more
runs per point in time you could actually relax this assumption and
use a 1...1 2...2 3...3 column vector instead.
Hope this helps
cheers
christian
On 25 Nov 2005, at 14:01, Isabell Wartenburger wrote:
> sorry, i hope it's in better format now? isabell.
>
> Dear experts,
>
> I have a treatment design with 3 time points (time1=before,
> time2=within,
> time3=after treatment).
>
> Each scanning consists of 2 functional runs (run1 and run2, always
> the same
> task/design with four different randomized tasks in each run).
> I am interested in the effect of the treatment on the tasks IN A
> SINGLE
> SUBJECT.
>
>
> #(1) I have done a first level analysis per time and run (first level
> within session) [result: 6 .feat folders (one per run and time)].
>
> #(2) I fed the .feat directories into three second level analysis
> per time
> (higher level analysis, inputs-are-lower-level-feat-directories, fixed
> effects) [result: 3 .gfeat folder (one per time)].
>
> #(3) I thought I should now fed the .feat directories of step (2) into
> another third level analysis (again higher level analysis, inputs-
> are-lower-
> level-feat directories) [result would be one .gfeat folder per
> contrast].
> Here I wanted to determine the effect of the treatment (e.g.
> Task1Time1 vs.
> Task1Time2; or Task1Time1 vs. Task1Time3 etc.).
>
>
> How do I set the contrasts for ONE SINGLE SUBJECT? (I know, for a
> GROUP I
> should use the "Tripled Two-Group Difference ("Tripled" T-Test)"
> example
> from the Feat-in-Detail-Homepage.)
>
>
> I've tried the following:
>
> Group EV1(time1) EV2(time2) EV3(time3)
> 1 1 0 0
> 1 0 1 0
> 1 0 0 1
>
>
> Contrasts:
> time1-time2 [1 –1 0]
> time2-time3 [0 1 –1]
> time1-time3 [1 0 –1]
>
>
> But after the Starting-higher-level-stats... the following error
> message
> occurs:
>
>
> ******************
> An exception has been thrown
>
> Singular matrix. Trace: Gsmanager::ols; Gsmanager::run.
>
> echo 0 > stats/dof
>
> /bin/rm -f stats/zem* stats/zols* stats/mask*
> Rendering using zmin=2.3 zmax=8
>
> mkdir tsplot
>
> /usr/local/fsl/bin/tsplot . -f filtered_func_data -o tsplot
> ** ERROR: nifti_image_read(./stats/pe1): can't open header file
> ** ERROR: nifti_image_open(./stats/pe1): bad header info
> Error: failed to open file ./stats/pe1
> ERROR: Could not open image ./stats/pe1
> Image Exception : #22 :: Failed to read volume ./stats/pe1
>
> Finished FEAT at Fr Nov 25 10:10:56 CET 2005
> To view the FEAT report point your web browser at
> /home/isa/data/berli/berli_07/fsl_test/
> fsl_mail_single_subject.gfeat/cope1.f
> eat/report.html
>
> ******************
>
>
> (I got the same error if I tried EVs and contrasts as in the
> "Tripled Two-
> Group Difference ("Tripled" T-Test)" example and if I tried to fed in
> Inputs-are-3D-cope-images-from-Feat-directories.)
>
>
> The error seems to be related to my settings of EVs and contrasts.
>
> Where can I find information on how to set the contrasts for single
> subject
> analyses?
>
> #(4) Or do I have to skip step (2) and take the 6 .feat folders of
> step (1)
> to the final analysis?
>
>
> Group EV1(time1) EV2(time2) EV3(time3)
> 1 1 0 0
> 1 1 0 0
> 1 0 1 0
> 1 0 1 0
> 1 0 0 1
> 1 0 0 1
>
>
> Contrasts:
> time1-time2 [1 –1 0]
> time2-time3 [0 1 –1]
> time1-time3 [1 0 –1]
>
>
> This works, but how does FSL then now that this is only one subject?
>
>
>
> Thank you so much.
> Best regards, isabell.
--
Christian F. Beckmann
Oxford University Centre for Functional
Magnetic Resonance Imaging of the Brain,
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
Email: [log in to unmask] - http://www.fmrib.ox.ac.uk/~beckmann/
Phone: +44(0)1865 222551 Fax: +44(0)1865 222717
|