Hi Gaby, we've been looking into this and have just noticed that your
design.mat for the 100 subejcts looks really odd - is it correct? I
think it probably explains your results...
Cheers, Steve.
On 12 Jul 2006, at 07:00, Gaby Pell wrote:
> Dear Steve and Jesper,
>
> Thanks for your responses.
>
> I tried changing the cluster correction threshold, and also tried
> redoing
> the analysis on a single slice - but I got a similar result of a
> uniform
> probability throughout the tbss maxc map.
>
> I uploaded the data to the FSL website (750457).
>
> Thanks,
>
> Gaby
>
>
>> On 10 Jul 2006, at 16:47, Jesper Andersson wrote:
>> Hi again Gaby,
>
>> there might be a "natural" explanation.
>
>> To me it sounds as if it might be a quirk of cluster based
>> stats. Let us say you have two groups and there is a general
>> decrease in
>> FA in one of the groups, and that this is such that all of the
>> sceleton
>> survives some lowish threshold (2.3). You will then end up with a
>> single
>> cluster that will have a uniform and very low probability.
>
>> What happens if you raise the initial threshold of the cluster-based
>> test to e.g. ~3.1?
>
>> If this is indeed the problem I think it demonstrates the (general)
>> issue of interpretability of cluster-based tests when using low
>> initial
>> thresholds.
>
> Good luck Jesper
>
>
>> On 3 Jul 2006, at 16:47, Gaby Pell wrote:
>>
>>> Dear Steve,
>>>
>>> I have been trying to run a tbss FA comparison between 2 groups
>>> with a
>>> large group of around 100 subjects in one group and around 20 in
>>> the other
>>> (a bit skewed I know!). I am finding though that after randomise,
>>> the
>>> cluster corrected threshold maps (maxc) don't look as expected. The
>>> whole
>>> skeleton is visible with a uniform probability value throughout.
>>> The voxel
>>> corrected and voxel non-corrected maps look as expected. When I
>>> remove
>>> some of the controls and make this group a bit smaller (I have
>>> tried up to
>>> around 65 controls), the cluster corrected map looks as expected.
>>> Do you
>>> have any idea what may be happening?
>>>
>>> Thanks,
>>>
>>> Gaby
>>>
>>> ***************************************
>>> Gaby Pell, PhD
>>> Brain Research Institute
>>> Melbourne, Australia
>>
>>
>> ---------------------------------------------------------------------
>> ---
>> ---
>> 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
>> ---------------------------------------------------------------------
>> ---
>> ---
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|