That's the problem with cluster-based methods - there's no good way
of choosing the cluster-forming threshold, or the data smoothing.
If your data is weak you could try reducing the -c threshold to even
as low as 1.5 or lower (this is valid in randomise, but not in GRF).
You could also try FDR - see the FSL manual page for this.
Cheers.
On 15 Mar 2007, at 01:36, bs Jeong wrote:
> Thank you for your recommendation.
> It sounds everybody can make something with the cluster-based
> correction.
> I was not able to find any significant change after either '-c 3'
> and '-c
> 2' option with 5000 times permutation.
> Please let me know what in tstat maps give me suggestion to decide
> adequate '-c value'.
> Could you see my t maps?
>
> thanks,
>
> bsJeong
>
>> Hi,
>>
>> I would definitely recommend to present some results
>> corrected for multiple comparisons, rather than going
>> for any voxel number threshold in your uncorrected
>> maps.
>> It looks like you have a lot of significant
>> uncorrected results for p<0.005, so I am quite
>> surprised you do not obtain anything with the
>> cluster-based correction.
>> What t threshold have you used in your randomise
>> option -c? You may have chosen a too stringent level.
>> I would therefore suggest to look at your tstat maps
>> to find out which -c would seem reasonable to enter...
>>
>> Best,
>> Gwenaelle
>>
>>
>> --- bs Jeong <[log in to unmask]> a écrit :
>>
>>> Dear FSLer
>>>
>>> I did TBSS analysis.
>>> But I did not any significant difference between
>>> groups in cluster level.
>>> So I'd rather find something in tbss_vox_stat with p
>>> value or tbss_tstat
>>> with t value.
>>> Problem is that so many regions are showed as
>>> difference btw group even
>>> though 1-p value is thresholded with 0.9949.
>>> Is there any method to threshold with consecutive
>>> voxel number?
>>>
>>> Thanks,
>>>
>>> bsjeong
>>>
>>
>>
>>
>>
>>
>>
>>
>> _____________________________________________________________________
>> ______
>> 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
>>
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|