Hi Marieke,
See this page:
http://fsl.fmrib.ox.ac.uk/fsl/fsl4.0/randomise/index.html
And consider using TFCE instead of using a cluster forming threshold like .001:
http://www.ncbi.nlm.nih.gov/pubmed/18501637
randomise also supports TFCE.
Also, see supplemental figure 7 here:
http://www.jneurosci.org/content/suppl/2010/02/12/30.6.2268.DC1/Supplemental_Material.pdf
Donna
On Oct 18, 2011, at 8:15 AM, Matthew Webster wrote:
> Hello,
> I'm not sure exactly what you're looking for, but the 'cluster' tool ( described in http://www.fmrib.ox.ac.uk/analysis/techrep/tr00df1/tr00df1/ ) can calculate p-values via cluster thresholding. Cluster-thresholding is also offered by randomise, our nonparametric statistical tool.
>
> Many Regards
>
> Matthew
>
>> Hi,
>>
>> I have a thresholding question:
>> Is it possible to estimate the false-positive rate corresponding to a given uncorrected p-value and minimum cluster size? Or, similarly, to estimate which minimum cluster size (number of voxels) is required to obtain a certain false positive rate (eg, .01) for a given uncorrected p-value (eg .001)?
>> AFNI and BrainVoyager have tools to calculate this, but I couldn’t find a way to do this in FSL. Does anyone know whether/how something like this is possible in FSL?
>>
>> Thanks!
>>
>> Marieke
>>
|