Hi,
On 10 Aug 2006, at 21:59, Dan Wolf wrote:
> Hello,
> A few questions/comments about randomise:
> 1. I also get strange masking effects if I do not supply a mask.
Yes - currently the rule is that if you don't supply a mask it takes
the mean of the data over time and thresholds at 0.001 to create the
mask - this will not be appropriate for all data (including data with
-ve values) - we'll probably change this in future to look at std()
instead. It's best to supply your own mask.
> 2. Am I correct that using the "-c" cluster option gives you an
> output file root_maxc_tstat where the
> value at each voxel in the image is the "cluster-corrected" p value
> for that voxel? In other words,
> "non-significant" clusters are shown, but will have values higher
> than .05 or whatever cutoff one
> wants to use?
Nearly. This does indeed give you the correctede p-value for that
cluster. However, as with the other p-value output from randomise,
the values output are actually 1-p for convenience of display,
thresholding etc. Hence if you want p<0.05 then just threshold at
0.95. If you have missing clusters then this is because they are
smaller than EVERY permutation's largest cluster.
> 3. The values in the root_max_tstat image do not make sense to me,
> in that they are very low (ie,
> largest is .08) so that they look like p values, but the numbers go
> UP as the tstat for that voxel goes
> up. My understanding is that they are p values, but then they
> should go down as one moves into a
> peak of an activation, which does not seem to happen.
That's right - that's because it's 1-p - it does say so in the manual!
http://www.fmrib.ox.ac.uk/fsl/randomise
Cheers.
>
> Thanks,
> Dan
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|