Dear Torsten,
you can find the relevant information on
http://www.fmrib.ox.ac.uk/fsl/randomise/index.html.
It will indicate you that max gives you the p-values
"VOXEL-based corrected" for multiple comparisons while
maxc gives you p-values "CLUSTER-based corrected".
Cluster-based thresholding happens to be much more
sensitive than any voxel-based one so I'm not
surprised about your results. Having said that, even
though this approach is valid, it would be better not
to use a too liberal threshold in your t-value
(randomise ... -c xx).
By thresholding your maxc map at 0.95, you will indeed
obtain all the clusters that are significant in your
analysis. You could further use them to mask the raw
t-map (tstatx) and report all the local maxima for
instance.
Hope this helps,
Gwenaelle
PS: I'm not good at statistics so my point of view is
mainly an "applicative" one!
> Dear all,
>
> I run randomise on a set of images, and was
> wondering about the output. I
> assume the _maxc_tstat, and _max_tstat images are
> the "important"
> statistics. I understand that the _max_tstat images
> describe a voxel by
> voxel comparison (is this right?), for which I
> basically do not get anything
> significant. On the other hand, I got a lot of
> significant "clusters" in the
> _maxc_tstat images, and was wondering what these
> mean. The input data images
> are registered FA images, but not processed with
> TBSS.
>
> In addition, the website reads that the values are
> 1-p (_maxc_tstat and
> _max_tstat images), so I would need to threshold
> those to see significant
> voxels. Would this mean that I chose the intensity
> minimum in fslview to be
> 0.94, and the max 0.95 to see only the significant
> voxels?
>
> Sorry for the rather basic questions and my neglect
> for stats, but for my
> excuse: I searched the website and couldn't find a
> describtion on how to
> interpret this data.
>
> Thanks in advance for any help regarding this issue.
>
>
> Torsten
>
___________________________________________________________________________
Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions !
Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses
http://fr.answers.yahoo.com
|