Print

Print
Hi Charlie,

Please, see below:

On 28 June 2014 01:14, charlie <[log in to unmask]> wrote:
FSL group,

Re the use of fdr with randomize, I fed the string below into fdr and the probability threshold below resulted.
-How is this number to be “applied” to the p-value image?


Use fslmaths to threshold the image. Something as:

fslmaths LgRD9_tfce_p_tstat1.nii.gz -thr 0.9921875 LgRD9_tfce_p_tstat1_fdr

where 0.9921875 is 1 - 0.0078125

 
-Also, the thresholded FDR image created seems much less stringent than the permuted corrp_stat image when thresholded in fslview.

Yes, FDR is less stringent than FWER correction. FDR means that, on average, 5% of the discoveries are false positives, whereas FWER means that, if the same experiment were repeated many times, in 5% a false positive would appear anywhere in the image.

All the best,

Anderson


 

fdr -i LgRD9_tfce_p_tstat1.nii.gz --oneminusp -m mean_FA_symmetrised_skeleton_mask.nii.gz -q 0.10 --othresh=thresh_LgRDanfdrstat1
Probability Threshold is:
0.0078125

Thank you