Hi,
There are two possibilities here I think.
The first is that in general an f-test may be more or less sensitive
than the things it is combining, as the degrees-of-freedom changes. In
general it is more sensitive but not always.
A second factor is that the FLAME FE option is not "OLS"; it uses the
nice weighting features that are also in the FLAME ME approach, so
that lower-level inputs get weighted according to their variances.
Maybe this explains the differences you are seeing?
Cheers.
On 21 Jan 2008, at 18:54, Jack Grinband wrote:
>> Are you talking about "two regressors" at the first or second level?
>> I'm not quite following what's happening yet.....
>
> There are two regressors in the model at the first level. I do a [1
> 0] and
> a [0 1] contrast and also an F-test. This is done for several
> functional
> runs. I then do a standard fixed effects analysis at the 2nd level
> for each
> regressor and a group analysis for the F-test using sum(Z)/sqrt(n).
>
> Finally, I count the number of significant voxels at the 2nd level
> for each
> regressor and compare with the number of significant voxels in the F-
> test.
> thanks,
>
> jack
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|