Hi,
> 2 quick question, i run an analysis using randomize with tfce, the correct p data from the analysis i transformed using fslmaths OneSampT_tfce_corrp_tstat1.nii.gz -log -div -2.3 -thr 1.3 output.nii.gz (thank you Anderson M. Winkler!).
>
> 1) i want to use -cluster function to list the clusters but i am not sure which number to use for thresholding, should i use the same value of 1.3 as in the transformation?
If you are using the output from the command above then any threshold value between 0 and 1.3 will give you the same answer (as the fslmaths command has set any values less than 1.3 to be zero). Choosing 1.3 is therefore sensible and report on all the super-threshold voxels.
> 2) I have quite a few very small clusters (smaller then 10 voxels, and as small as 1 voxel), is there a guideline to what cluster size can be used, and what size should be overlooked?
If you started with a corrected p-value image from randomise, then multiple comparison correction has already been done and hence clusters of all sizes are statistically valid. However, biologically small clusters are hard to justify, but I don't think there is an agreed threshold value. I would report all your clusters and then comment on this issue in the discussion.
All the best,
Mark
>
> Thank you
>
> Navot
|