Hi Chris,
For others reading this, Chris is referring to the example at:
http://www.fmrib.ox.ac.uk/fsl/feat5/detail.html#TripledTwoGroupDifference
On Fri, 9 Sep 2005, Chris Kelly wrote:
> Hello,
>
> I am reading the instructions for performing a tripled two-group
> difference analysis using FEAT.
>
> The concept of demeaning each subject makes perfect sense. Also, I
> follow the algebra for finding A- C and so on. What I don't understand,
> however, is how one makes the initial assignments (A=a+b, B=-a and C=-b)
> that get used to find A-C. Nor do I understand how EV1 and EV2 were
> initially designed.
OK - what we're saying is that, due to the subject-wise demeaning, the
responses to the three conditions A, B and C must add up to zero - i.e.
have zero mean across the different conditions. In the example, the first
5 scans on average contain the level of condition A in the data (relative
to this zero mean across conditions), the next 5 scans comprise condition
B and the final five are condition C.
Hence the first 2 EVs, which both have value 1 for the first 5 scans,
between them model condition A: so after estimating the PEs, if we define
PE1=a and PE2=b, the modelled data for those first 5 scans is 1*a + 1*b,
i.e., A=a+b.
Likewise, for the next five scans, EV2 is 0, and EV1 is -1, hence the
modelling of the data is: B=-a. Likewise, C=-b.
Hopefully this clarifies things?
> Also, I'm not sure why the following setup would not work:
>
> EV1=[1 1 1 1 1 0 0 0 0 0 0 0 0 0 0]
> EV2=[0 0 0 0 0 1 1 1 1 1 0 0 0 0 0]
This would not work because there is now nothing in the model to model
condition C, hence there can't be a good model fit!
Cheers, Steve.
--
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
|