Hi - I suspect that either the difference lies in the exact equation
that you're using (see for example eq 14.9 in Keith's chapter in our
FMRI book), or some implementational detail, in which case Wooly will
need to help here.
Cheers, Steve.
On 16 Oct 2007, at 05:33, Jeanette Mumford wrote:
> Hi,
>
> I was just trying to recreate the varcope image from a first level
> analysis using the sigmasquareds and threshac1 images as well as the
> design.mat file. My result is close, but I'm wondering why it isn't
> closer. Here's what I've done
>
> 1. Create the correlation matrix V using the first element of
> threshac1 as the main diagonal, 2nd element as the first diagonal
> above and below the main diagonal and so forth.
> 2. Demean the design matrix that's stored in design.mat (call it X)
> 3. Calculate inv(X'*inv(V)*X)*sigma^2 Where sigma^2 is the
> sigmasquareds value for that voxel and take the appropriate contrast
> of this matrix to correspond with my varcope of interest.
>
> My method seems to produce a variance that is consistently smaller
> than the varcope, but within 20% or so. Did I miss something?
> Highpass filtering was used in the analysis (both the data and the
> design), but I figured that would already be incorporated in all the
> pieces that I used. The analysis was originally carried out in Feat
> version 5.43.
>
> Thanks!
> Jeanette
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|