Isn't the impact of the prewhitening already included in the varcope
value?
To the original Stephen (Towler): It isn't clear to me why you are
getting such a "big" difference. Did you do the comparison directly on
the images, using fslmaths, so that you didn't introduce your own
rounding errors?
For example, for a Level 1 analysis that included FILM prewhitening, I
did the following:
fslmaths varcope1.nii.gz -sqrt -div cope1.nii.gz -recip mytstat1
fslmaths tstat1 -sub mytstat1 diff
fslstats diff -R
In the case that I tested this on, the resulting range of 'diff' was
-0.000000 0.000000
indicating that tstat indeed exactly equals cope/(sqrt(varcope)) (up to
a trivial precision difference, which is probably due to the fact that
the fslmaths code that I used above to compute 'mytstat1' differs
slightly in the order of operations relative to what FEAT is using
internally.
cheers,
-MH
On Fri, 2011-04-01 at 08:06 +0100, Stephen Smith wrote:
> Hi - it also prewhitens the data - see the Woolrich Neuroimage paper
> on that.
> Cheers.
>
>
>
> On 31 Mar 2011, at 18:06, Stephen Towler wrote:
>
> > Hi authors and experts,
> >
> > Computationally, how is FEAT calculating each PE's t-stat? The
> > manual
> > says tstat=cope/(sqrt(varcope)), but that doesn't quite work when I
> > do
> > the arithmetic manually. At one example voxel, the FEAT tstat is
> > 3.03759, while cope/(sqrt(varcope)) = 0.130471/(sqrt(0.0018)) =
> > 3.0752. The difference may be decimal dust, but it does pique my
> > interest in how FEAT tstat was computed, especially given that
> > entering the same the same x and y data into a SPSS simple OLSQ
> > regression results in a PE identical to FEAT's, but a third, much
> > lower, tstat (2.983).
> >
> > Thanks!
> > Stephen
> >
> >
>
>
> ---------------------------------------------------------------------------
> 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
> ---------------------------------------------------------------------------
>
>
>
>
>
>
>
|