OK,
So let's say we have used 3dFDR to calculate a statistical map of q values after its ranking algorithm has had its way with some statistical dataset-- such as a group map of "activations" as detected by a linear contrast.
Could we (i.e. do users) then use this new dataset output as a mask to create a "final" dataset for analysis and reporting which has filtered out from consideration voxels whose absolute q value was below the two-tailed threshold required for p < .05?
An example would be thus:
If 3dDFR were used on the "ConditionA_vs_ConditionB_tlrc" groupwise statistical map
3dFDR -input ConditionA_vs_ConditionB_tlrc[1] -mask_file TLRCgroupmask+tlrc -cdep -prefix FDR_AvsB
to make a map of q-scores "FDR_AvsB"
Would it "be reasonable" for me to then use 3dmerge to make a nice (filtered) map of actual voxel-wise p-values for reporting in a publication, where the possibility of an erroneously-reported "active" voxel has been reduced to P < 0.05?
3dmerge -prefix AvsB_FDRcontrolled -1thresh 1.96 'ConditionA_vs_ConditionB_tlrc[1]'
where 1.96 is the critical value for p < .05 in a two-tailed distribution.
Jim B