Dear AFNI experts,
I am looking for some advice regarding process the 3dClustSim (forgive me if my question is too basic…).
I tried to conduct AlphaSim, and I could find that 3dClustSim is one of the method to measure it.
However, I also found the another way to measure.
Input an option ‘-Clustsim’ during conducting 3dttest++
3dttest++ -prefix prefix_name -AminusB -paired -Clustsim
3dClustSim -nxyz 61 73 61 -dxyz 3 3 3 -both -prefix prefix_name -pthr 0.05 0.01 0.005 0.001 -athr 0.05 0.01
The question is, when I conduct those two methods separately, I could get slightly different results.
But, I’m not sure whether I used 3dClustSim function correctly, so it is difficult for me to find out which one is correct or better method to use…
Thus, may I ask you which method that I could select to get AlphaSim result?
Thank you in advance…
AlphaSim, aside from being subsumed by 3dClustSim, is no longer
recommended because FWHM estimates are not considered accurate
measures of spatial correlation. Using 3dFWHMx and 3dClustSim
with the ACF method is considered much more accurate.
Please see the Clustering in AFNI post, above.
The 3dttest++ -Clustsim method is more of a permutation test.
In that it is different from random noise Monte Carlo simulations
done in 3dClustSim.
3dClustSim should always be run with the -mask option, rather
than -nxyz and -dxyz. Using -pthr of 0.05 and 0.01 might be too
high. It should be passed accurate ACF parameters via -ACF.
Either of these methods should be valid, but method #2 depends
on getting accurate blur estimates to pass via the -ACF option.
If, for example, AFNI is not being used to generate residuals
from which the ACF parameters are computed, we would not know
enough to be able say whether the method seems valid.
But method 1 should be more robust in that regard.
Thank you so much for the comments.
It’s really helped me to understand and conduct AlphaSim (-Clustsim).
Thank you again!