Dear Steve / FSL,
I have tried running the fdr command on my uncorrected p-value map (after
inverting it) as suggested and it always spits out a Probability Threshold of 0.
This happens whether I use the GM_mask or my frontal ROI mask as the mask
and also when I change the p-value from 0.05 to something more or less
conservative. Any ideas on how to get a more sensible p-value for
thresholding using FDR?
Many thanks,
Natalia
>On 11 Aug 2008, at 14:30, Natalia Lawrence wrote:
>
>> In particular, I am trying to replicate a
>> published structural VBM analysis that used SPM5 and a non-parametric
>> approach but thresholded at p < 0.05 corrected for multiple
>> comparisons
>> following a False Discovery Rate methodology; is there an equivalent
>> in FSL-
>> VBM?
>
>Yes, you can feed the randomise-output uncorrected p-values into FDR
>thresholding - see the FDR manual for an example on that:
>http://www.fmrib.ox.ac.uk/fsl/randomise/fdr.html
>
|