Dear all,
I have a test-dataset containing RS-data which I can use to look at the effect of different analysis methods. I have run Randomise on this using:
randomise_parallel -i dualregression_ic0000.nii.gz -o randomise_out/output -d design.mat -t design.con -f design.fts -e design.grp -n 500 -m melodic/mask.nii.gz -T -x --uncorrp
after this, I ran FDR on the tfce_p outputs from randomise:
fdr -i randomise_tfce_p.nii.gz -q 0.05 --oneminusp -a output_FDR
I would assume that the two output maps (FWE and FDR) would be very similar, with the FDR being slightly "bigger", as it would gain 5% false positives over de FWE-corrected maps. However, this is not the case and the output maps are very different. There are significant differences in the tfce_corrp-maps, but not in the tfce_p_FDR-corrected ones.
Is my line of thinking incorrect or am I doing something wrong here?
Kind regards,
Dennis
|