Dear Tim, Doug and Steve,
Thanks for your answers, this puts me on the right track in understanding
and applying non-parametric cluster thresholding!
regards,
Bas
-----Oorspronkelijk bericht-----
Van: FSL - FMRIB's Software Library [mailto:[log in to unmask]]Namens
Steve Smith
Verzonden: vrijdag 17 februari 2006 18:41
Aan: [log in to unmask]
Onderwerp: Re: [FSL] Cluster thresholding using randomise
Yo Doug,
On 17 Feb 2006, at 16:53, Doug Greve wrote:
> Yes, c_thresh is a t-value.
>
> You don't need to specify a cluster size threshold -- this is what
> randomise is computing, ie, the probability of gettting a cluster
> of certain size or larger (this is the cluster-wise pvalue). Thus,
> by choosing a cluster-wise pvalue threshold, you are effectively
> choosing a cluster size threshold (but the pvalue is more
> meaningful:).
>
> A few more hints from my experience:
>
> I think randomise will try to compute a mask from the data if you
> don't pass one on the command-line. Is this right? If so, it's
> likely that it will not compute mask that make sense.
It should work in general - it takes the temporal mean of the input
data and thresholds that at 0.0001 to create the mask. But yes it is
safer to create exactly the right mask and pass that in.
Cheers, Steve.
> Tom Nichols recommends using 10,000 iterations to test at p <= .05
>
> The "neighborhood" of a voxel (ie, it's connected components) are
> the 26 voxels around it (touching edges and corners are connected,
> not just touching faces).
>
> Hope this helps!
>
> doug
>
>
> bas Jasperse wrote:
>
>> Dear FSL-list,
>>
>> Recently we started using randomise to perform voxelwise
>> statistics on SIENA displacement maps and are currently performing
>> correlations and group comparisons.
>>
>> We also want to use cluster thresholding for our analysis, but are
>> a bit puzzled by the option provided.
>>
>> The manual tells us that the only thing we can specify is
>> <c_thresh>. We assume that this means setting the supra-threshold
>> based on the
>> T-stat image. Is this correct?
>>
>> There is no option provided for cluster size in voxels or mm, so
>> we were wondering
>> if and how randomise sets the cluster size. Could you provide some
>> insight in this matter?
>>
>>
>> We've tried searching the Jiscmail archives, but could not find an
>> answer to these issues.
>>
>> Thanks in advance for your help,
>> kind regards,
>>
>> Bas Jasperse
>> Vu medical centre, Amsterdam
>>
>>
>>
>
> --
> Douglas N. Greve, Ph.D.
> MGH-NMR Center
> [log in to unmask]
> Phone Number: 617-724-2358 Fax: 617-726-7422
>
> In order to help us help you, please follow the steps in:
> surfer.nmr.mgh.harvard.edu/fswiki/BugReporting
------------------------------------------------------------------------
------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, 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
------------------------------------------------------------------------
------------------
----------------------------------------------------------------
VU medisch centrum. Kennis maakt ons beter.
Lustrum 20056
www.5jaarvumc.nl
|