Aha - Tim's kind of right, but forgetting something important - in fact,
running ICA (MELODIC) on the residuals of the FEAT analysis (at first or
second level) is potentially even more powerful than Tom's univariate
investigation of non-Gaussianity; it is explicitly looking for
spatiotemporal patterns of structured noise in the data (in this case the
residuals). It gives you a new report of the spatial and temporal
characteristics of each structured process, after estimating automatically
the number of processes present.
To run this, you will need a minor edit on fsl/tcl/feat.tcl so that the
residuals don't get cleaned up (to save disk space); find the line that
removes res4d.hdr and res4d.img and comment it out. Then find
.feat/stats/res4d and feed this into the Melodic GUI.
Cheers, Steve.
On Thu, 6 May 2004, Clark Johnson wrote:
> Greetings:
>
> I've been looking for a more extensive set of tools to discover outliers and
> violations of underlying regression assumptions. I have not located much in
> the FSL world but did find a very exciting SPM99 extention called SPMd -- a
> regression diagnostic toolbox (www.sph.umich.edu/~nichols/SPMd).
>
> 1) Have I missed the FSL counterpart to this set of tools?
>
> 2) If not, I was wondering if anyone has ported SPMd to an FSL enviroment?
>
> 3) Perhaps you guys (e.g. the FSL superstars) have something like this "in
> the works." If so, then I'll save myself the headache of trying to figure
> this out.
>
> 4) On the other hand -- perhaps these methods are not appropriate for the
> statistical methods used in FSL.Any thoughts?
>
> Tnx
>
> Clark
>
> O
> o
> ||||| o
> - -
> @(0)-(0)@
> ----------oOOO-----U-----OOOo---------------
> Clark Johnson
> Research Associate Professor
> Box 357263
> Seattle, WA 98195-7263
> P:(206)685-0818 F:(206)685-9551
>
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
|