Hi
I'll give it a try.
EV1/EV2 are entirely orthogonal to EV3-EV7 - therefore, the final 5
EVs only model the subject-specific mean and have not impact on the
parameters associated with EV1 and EV2 (a and b). These final EVs
model all the signal mean, while EV1 and EV2 model the variations
around the mean under the different conditions.
a and b are the fitted parameter estimates from EV1 and EV2, i.e.
describe the amplitude of the overall signal change detected in the
GLM. Therefore, any signal change detected under condition A is equal
to a+b (that's the total 'height' modelled). Similarly any signal
change away from the mean under condition B is modelled by -a and any
change away from the mean under condition C is -b, therefore
A=a+b
B=-a
C=-b
Incidentally it is then the case that a + b = (2A - B - C) / 3, i.e.
it is indeed the case that signals changes under all three conditions
contribute equally (1/3) to fitting the parameters a and b and your
first assumption was almost bang on.
The remainder of what's given on our webpages then follows naturally
from the equations above, i.e. if we're interested in where the
signal under condition A exceeds the signal under conditon C we need
the 'A-C>0' contrast with it's estimate being a+b-(-b), i.e we need
the [1 2] contrast
hope this helps
Christian
On 23 Jul 2007, at 21:31, Sophie Anisa wrote:
> Hello
>
> I wonder whether some kind soul could explain to me in a way that I
> can understand
> the logic of the tripled t-test that is shown on the FSL help page:
>
> http://www.fmrib.ox.ac.uk/fsl/feat5/
> detail.html#TripledTwoGroupDifference
>
> I can sort of see what is meant to be going on but sadly I'm just
> not getting it when it comes to the statement:
>
> > We define PE1=a and PE2=b. Then, we can see by looking at the three
> > condition blocks, that the mean ... of condition A is modelled by
> A=a+b
> So... why does this correctly model the mean of condition A?
> It looks like a+b should model 2A-B-C?
>
> Or if, instead, I set S=PE3+PE4+PE5+PE6+PE7
> It looks like A=(a+b+S)/3?
>
> What am I missing?
>
> kind regards
>
> Sophie
____
Christian F. Beckmann
University Research Lecturer
Oxford University Centre for Functional MRI of the Brain (FMRIB)
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
[log in to unmask] http://www.fmrib.ox.ac.uk/~beckmann
tel: +44 1865 222551 fax: +44 1865 222717
|