That is correct.
The mathematics used for the clustering inference require the
simple definition of cluster that we've implemented.
With randomise other cluster definitions are possible (although
not readily available via our code), but the new TFCE method
also does clustering differently, so these are alternative options.
All the best,
Mark
Michal Kuniecki wrote:
> Hello Benny,
>
>
> Funny but if I remember well I posted very similar question to the
> list here:
>
> http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind06&L=FSL&D=0&X=0BA7EA71C6B9
> 7C5879&Y=kuniecki%40gmail.com&P=97392
>
> And if I understood answers to my question good enough,
> unfortunately there is no way to affect clustering other then manipulating
> the threshold. This is quite a pity actually as there are methods for
> determining independent clusters which are used in microbiology for example.
>
>
> OK, this is only a hint, I am by no means an expert in FSL.
>
> Mike
>
>
>> -----Original Message-----
>> From: FSL - FMRIB's Software Library
>> [mailto:[log in to unmask]] On Behalf Of Benny Liberg
>> Sent: Thursday, July 31, 2008 10:54 AM
>> To: [log in to unmask]
>> Subject: [FSL] delimitation of cluster
>>
>> Hi,
>>
>> I have done a second level analysis in FEAT (voxel corrected)
>> and I use the cluster program to get an unbiased report of
>> the co-ordinates of local maximas in my zstat images.
>>
>> Functionally different premotor areas (in my case SMA, dorsal
>> PMC, ventral
>> PMC) are reported as one cluster since the PMC cluster in the
>> thresh_zstat image is continous. However, there are very few
>> voxels making it one continous cluster and not three smaller
>> ones, which is how we would expect it to be. The cluster
>> seperates into three if I change the threshold to 0.01, but I
>> would like to keep it at 0.05 for comparison reasons.
>>
>> Is there any way that I can affect the cluster delimitation
>> in the cluster program?
>>
>> Regards,
>> Benny
>>
>>
>
>
