Hi, this is answered at:
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0711&L=FSL&P=R22957&I=-3&X=7255C94DDF245A3414&Y=steve%40fmrib.ox.ac.uk
Cheers.
On 2 Mar 2008, at 21:01, Daniel Irlam wrote:
>
> Hi All
>
> I'm confused by another issue regarding PPheights after running a
> recent
> Featquery ROI analysis on a single subject (single .feat directory).
> The design
> was very simple - a block design with only three conditions (A,B,C)
> alternating
> with rest. In the FEAT design, I put condition A as Event 1, B as
> Event 2 and C
> as Event 3. Temporal derivatives were included as regressors of no
> interest.
>
> As I understand, this means that PE1 models A, PE3 models B and PE5
> models C.
> I also specified as my contrasts of interest:
>
> A 1 0 0
> B 0 1 0
> C 0 0 1
>
> When I came to run Featquery, as a sanity check, I specified both
>
> stats/pe1 stats/pe3 stats/pe5 and also
> stats/cope1 stats/cope2 stats/cope3
>
> (and clicked the button for "Convert PE/COPE values to %")
>
> I was assuming the results would be the same for the two rows.
> But they are not:
>
> 1 stats/pe1 371 -1.17 -0.1867 0.4408 0.3738 1.289 1.837 48 45 24
> -51.3 15.2 24.9
> 1 stats/pe3 371 -1.459 -0.2462 0.3588 0.3036 1.135 1.817 46 40 24
> -43.1 1.1 22.8
> 1 stats/pe5 371 -1.444 -0.1379 0.4546 0.3737 1.323 2.049 47 41 24
> -46.6 3.7 23.2
> 1 stats/cope1 371 -1.005 -0.1603 0.3785 0.3209 1.106 1.577 48 45 24
> -51.3 15.2 24.9
> 1 stats/cope2 371 -1.451 -0.2449 0.3569 0.302 1.129 1.807 46 40 24
> -43.1 1.1 22.8
> 1 stats/cope3 371 -1.267 -0.1209 0.3987 0.3277 1.161 1.797 47 41 24
> -46.6 3.7 23.2
>
> It appears the difference arises from the PPheights values being
> different in design.mat and design.con:
>
> design.mat:/PPheights 8.995116e-01 6.795912e-01
> 8.995116e-01 6.795955e-01 8.995116e-01 6.793940e-01 ...
> design.con:/PPheights 7.723186e-01 8.948356e-01
> 7.889021e-01
>
> So for example the max % signal change for PE1 (1.837) can be
> calculated from the max % signal change for COPE1 (1.577)
> and the ratio of the two PPheights values:
>
> 1.837 = 1.577 * (8.995116e-01 / 7.723186e-01)
>
> I'm puzzled by this as intuitively I had expected them to be the same?
> Indeed the images themselves (pe1.nii.gz and cope1.nii.gz *are*
> identical)
> so the % difference is driven entirely by the PPheights difference.
>
> Questions:
>
> Why are the PPheights different for the PE1 and the (1,0,0) COPE1
> contrast?
>
> Given they are different which measure - the %PE1 or the %COPE1 - is
> the more accurate estimate of the signal
> change for condition A in this subject?
>
>
> many thanks
>
> --
> Daniel Irlam Ph.D
> [log in to unmask]
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|