Im trying to run 3dClustSim -acf a b c -mask Mask.nii.gz -pthr 0.05 0.01 0.001 -iter 10000 -fwhm X
I would like to have some help deciding the number that I should use as my blur (fwhm). My preprocessed data have already a 5mm blur, should I use the same here for the 3dClustSim command?
I am also confused and I would like to have some clarification. Im not understanding correct if I should use or not the -fwhm flag in the 3dClustSim
In the webpage it says:
**** NOTICE ****
You should use the -acf method, NOT the -fwhm method, when determining cluster-size thresholds for FMRI data. The -acf method will give more accurate false positive rate (FPR) control.
-acf a b c = Alternative to Gaussian filtering: use the spherical
autocorrelation function parameters output by 3dFWHMx
to do non-Gaussian (long-tailed) filtering.
* Using ‘-acf’ will make ‘-fwhm’ pointless!
* The ‘a’ parameter must be between 0 and 1.
* The ‘b’ and ‘c’ parameters (scale radii) must be positive.
* The spatial autocorrelation function is given by
ACF(r) = a * exp(-rr/(2bb)) + (1-a)exp(-r/c)
[u][u]>>---------->>* Combined with 3dFWHMx, the ‘-acf’ method is now a
recommended way to generate clustering statistics in AFNI![/u][/u]
But the last part reffers to the combination with the 3dFWHMx
The 3dClustSim docs also say, under the -fwhm option, “This option is no longer recommended, …”. I think that unless you’re trying to do something special that falls outside the usual multiple-comparison workflow, and you know what you’re doing, you shouldn’t use that option.
What are you trying to accomplish with 3dClustSim? I’ve only ever used it in conjunction with 3dFWHMx.
Hi Andrew thanks for your answer. Im trying to compare the brain activation of 2 groups of patients (low grief symptoms and high grief symptoms) during a emotional processing task. I ran 3dttest with the -ClistSim and -ETAC options but it seems that any voxels are surviving the corrections. I was thinking that the power to detect differences are small considering that I am not comparing the diseased with normal subjects so I wanted to check if with a least stringent threshold I could detect any differences between this groups and that is why I was trying to run 3dClustSim. Do you have any advices on another method that I can use to compare my groups?
Thank you so much and Im sorry if Im saying not sense things, Im pretty new in the fMRI field.
It sounds like you do want to use 3dClustSim, but you first need to run 3dFWHMx on your first-level residual data. And use the acf options, not fwhm. See Andy’s intro, for example: https://andysbrainbook.readthedocs.io/en/latest/fMRI_Short_Course/fMRI_Appendices/Appendix_A_ClusterCorrection.html#cluster-correction
On the other hand, if you’re just exploring your pilot data results, I recommend just opening your uncorrected stat maps in a viewer and thresholding manually.
There are 3 related methods for cluster correction, which might be causing confusion here.
Note that -Clustsim and -ETAC are different (but related) methods. Using both options would simply be for educational purposes, as only one method would be used for the cluster correction. Of course, the method that will be applied (including the uncorrected p-value threshold) should be chosen before viewing the results.
But those options do not require ACF parameters, and they do not require running 3dFWHMx. 3dttest++ with -Clustsim or with -ETAC will generate random clusters using permutation testing on the residuals of the t-test, so an estimate of the mixed model auto-correlation function (ACF) parameters is not needed.
Alternatively, if you were using afni_proc.py and estimated the blur (ACF params) in the residuals, those blurs could be applied using 3dClustSim to do the correction. But that is a different method from using 3dttest++ with -Clustsim or with -ETAC.