Hi,
On 11 Feb 2008, at 16:55, David Kideckel wrote:
> I have almost completed a TBSS analysis, and have several questions
> regarding certain functions in Randomise and elsewhere:
>
> 1) What exactly will the cluster mass comparison tell me vs. the
> cluster size. Will either give me any more meaningful information?
Instead of just using the number of voxels in the cluster as the test
statistic, it uses the number weighted by the height (or t-statistic
value). It has been claimed that this can increase sensitivity to many
kinds of blobs....but in my experience doesn't make much difference
compared with cluster size.
> 2) What is a typical F threshold that one can use when cmparing
> either 2 or 3 groups based solely on FA. For a - C value, it is
> recommended to use 1.5 or 3.
See previous posts on how to choose the cluster-forming
threshold......I'm afraid there is no objective way to choose this.
This is the primary reason why we came up with the TFCE method - I
suggest you try that instead of cluster-based thresholding in randomise.
> 3) What typical value of - V should be used if variance smoothing is
> used? What benefit (if any) will the -V option give?
Again, this is up to you. This conditions the variance estimation by
pooling over space. Typically you might start by trying a size of
order of a voxel or two.
> 4) Is the TFCE option recommended over the -C option in general? If
> so, why? If no, why not?
Ah, I got ahead of myself. Yes, try it. See
http://www.fmrib.ox.ac.uk/analysis/research/tfce/tfce-hbm07-poster.pdf
for more details.
> 5) Is 7 subjects in 1 group vs 12 and 11 in two other groups be
> sufficient to use the randomise option?
Yes.
> Is there a manual for FSL which would answer these questions in
> detail. I have scrutinized the relevant FSL documentation on the
> website, but have not been satisfied with the answers received.
I guess your questions are more about understanding the ideas behind
the analysis rather than the mechanisms of running the software tools,
which is mostly what the manuals cover. Maybe you should sign up to go
on the FSL course in Australia ;-)
Cheers.
>
>
> Thanks!
>
> David Kideckel, PhD Candidate
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|