Hi all,
I noticed today while running some analyses that the z-statistics produced when I run 3dttest++ with -Clustsim differ very slightly from those produced when I use -ETAC instead. I must be misunderstanding, but I expected both to be the same up to when the permutations start. I’ve pasted my 3dttest++ command below. The only difference between the two commands is using -Clustsim or -ETAC. I’ve also pasted the first part of the output for each test below. You can see the warning about q stats differs slightly as well. What produces the small difference?
Thank you for your help!
Anthony
3dttest++ -setA S-H
sub001 stats.sub.21750_REML+tlrc.‘[71]’
sub002 stats.sub.21767_REML+tlrc.‘[71]’
sub003 stats.sub.21824_REML+tlrc.‘[71]’
sub004 stats.sub.21826_REML+tlrc.‘[71]’
sub005 stats.sub.21827_REML+tlrc.‘[71]’
sub006 stats.sub.21838_REML+tlrc.‘[71]’
sub007 stats.sub.21840_REML+tlrc.‘[71]’
sub008 stats.sub.21843_REML+tlrc.‘[71]’
sub009 stats.sub.21847_REML+tlrc.‘[71]’
sub010 stats.sub.21858_REML+tlrc.‘[71]’
sub011 stats.sub.21862_REML+tlrc.‘[71]’
sub012 stats.sub.21866_REML+tlrc.‘[71]’
sub013 stats.sub.21873_REML+tlrc.‘[71]’
sub014 stats.sub.21875_REML+tlrc.‘[71]’
sub015 stats.sub.21876_REML+tlrc.‘[71]’
sub016 stats.sub.21878_REML+tlrc.‘[71]’
sub017 stats.sub.21889_REML+tlrc.‘[71]’
sub018 stats.sub.22100_REML+tlrc.‘[71]’
sub019 stats.sub.22104_REML+tlrc.‘[71]’
sub020 stats.sub.22115_REML+tlrc.‘[71]’
sub021 stats.sub.22122_REML+tlrc.‘[71]’
sub022 stats.sub.22135_REML+tlrc.‘[71]’
sub023 stats.sub.22144_REML+tlrc.‘[71]’
sub0024 stats.sub.21730_REML+tlrc.‘[71]’
-prefix Shift_gt_Hold_etac -ETAC -mask group_mask+tlrc.HEAD
++ 3dttest++: AFNI version=AFNI_20.1.17 (Jun 19 2020) [64-bit]
++ Authored by: Zhark++
++ option -setA :: processing as LONG form (label label dset label dset …)
++ have 24 volumes corresponding to option ‘-setA’
++ Number of 3dXClustSim threads set to 56
++ 51322 voxels in -mask dataset
++ === ETAC memory requirements:
- = 1,026,440,000 (1.0 billion) bytes of pseudo-data in temporary .sdat files.
- = It is best to store these files on a solid-state disk (SSD)
- = using the ‘-tempdir’ option.
- === ETAC will also use additional memory for the cluster tables
- = which is hard to predict, but can be as much as the amount
- = listed above.
- === There are 103,079,215,104 (103 billion) bytes of memory on your system.
++ loading -setA datasets
++ t-testing:0123456789.0123456789.0123456789.0123456789.0123456789.!
++ ---------- End of analyses – freeing workspaces ----------
++ Creating FDR curves in output dataset
*+ WARNING: Smallest FDR q [1 S-H_Zscr] = 0.354566 ==> few true single voxel detections - Added 1 FDR curve to dataset
++ Output dataset ./Shift_gt_Hold_etac+tlrc.BRIK
++ Output dataset ./Shift_gt_Hold_etac.resid.nii
++ 3dttest++: AFNI version=AFNI_20.1.17 (Jun 19 2020) [64-bit]
++ Authored by: Zhark++
++ option -setA :: processing as LONG form (label label dset label dset …)
++ have 24 volumes corresponding to option ‘-setA’
++ Number of -Clustsim threads set to 56
++ 51322 voxels in -mask dataset
++ loading -setA datasets
++ t-testing:0123456789.0123456789.0123456789.0123456789.0123456789.!
++ ---------- End of analyses – freeing workspaces ----------
++ Creating FDR curves in output dataset
*+ WARNING: Smallest FDR q [1 S-H_Zscr] = 0.293434 ==> few true single voxel detections
- Added 1 FDR curve to dataset
++ Output dataset ./Shift_gt_Hold_clustsim+tlrc.BRIK
++ Output dataset ./Shift_gt_Hold_clustsim.resid.nii