Hi!

I have been reading about how to obtain the final table of p-values and cluster sizes such that family wise errors are taken into account. I realize that I need to use 3dClustsim in combination with 3dFWHMx: This seems to be the latest bid. And I know this is a work in progress, much like is suggested by Bob's reply in this thread:

[afni.nimh.nih.gov]

So I know things can change. But, I would just like to confirm that the following apparoach is the currently reccomended one:

Let's say that I have 10 subjects. Then I need to get smoothness estimates from all of these 10 subjects. I run:

3dFWHMx -detrend -ACF NULL (no output file) -input errts.subX+tlrc -mask mask_group+tlrc. (found in subjX.results)

This gives a lot of output but I'm only interested in the ACF model parameters. E.g. 0.453301 3.45469 12.7055 (Are these reasonable for one subject?). The input can be the errts file, right? Or should it be a specific contrast?

Then I wonder: Should I do this for all 10 subjects and then average the resulting parameters? Is there a simple way to extract the parameters? The outputfiles has a lot of output and not simply these parameters. Any tip for simpel extraction? Should I use some | grep command perhaps?

Let's say I now have the 3 average parameters from my 10 subjects. Then I should put these into 3dClustsim. Is this the way to do it?

3dClustSim -mask TT_N27+tlrc (resampled to group_mask) -acf (parameters from above)

Are the results from 3dClustSim meant to be uniqe for different contrasts? In that case I'm doing something wrong I guess. I know you can use some output (some kind of residuals) from 3dttest++ (uniqe for each contrast) as input to 3dFWHMx but in the thread linked to above the errts file is used (which is not contrast specific).



robka95@batmobile:/data/dsk1/analysis/clustsim$ 3dClustSim -mask resample+tlrc. -acf 0.453301 3.45469 12.7061
++ 3dClustSim: AFNI version=AFNI_16.0.05 (Feb 10 2016) [64-bit]
++ Authored by: RW Cox and BD Ward
++ 33511 voxels in mask (30.80% of total)
++ ACF(0.45,3.45,12.71) => FWHM=10.60 => 46x55x43 pads to 80x80x64
++ Startup clock time = 0.1 s
++ Using 12 OpenMP threads
Simulating:0123456789.0123456789.0123456789.0123456789.012345678!
++ Clock time now = 62.6 s
# 3dClustSim -mask resample+tlrc. -acf 0.453301 3.45469 12.7061
# 1-sided thresholding
# Grid: 46x55x43 3.50x3.50x3.50 mm^3 (33511 voxels in mask)
#
# CLUSTER SIZE THRESHOLD(pthr,alpha) in Voxels
# -NN 1 | alpha = Prob(Cluster >= given size)
# pthr | .10000 .05000 .02000 .01000
# ------ | ------ ------ ------ ------
0.050000 430.7 546.2 717.5 833.0
0.020000 127.0 161.3 208.7 250.7
0.010000 62.2 79.0 104.1 127.0
0.005000 34.1 43.8 57.8 69.2
0.002000 17.4 22.3 29.1 34.9
0.001000 11.4 14.3 18.9 22.8
0.000500 7.7 9.9 12.8 15.8
0.000200 4.8 6.2 8.2 9.6
0.000100 3.5 4.5 5.9 7.2




Thanks in advance!

Best,
Robin



Edited 9 time(s). Last edit at 06/01/2016 09:47AM by Robin.