Question about the alphasim correction

Submitted by liufeng on

Dear experts,

I have a question about the paper "Cluster failure: Why fMRI inferences for spatial extent have inflated false-positive rates" and Cox's paper  "AFNI and Clustering: False Positive Rates Redux". 

In the first paper, they mentioned that "First, AFNI estimates the spatial group smoothness differently compared with SPM and FSL. AFNI averages smoothness estimates from the first-level analysis, whereas SPM and FSL estimate the group smoothness using the group residuals from the general linear model (32). The group smoothness used by 3dClustSim may for this reason be too low". 

I did not fully understand these sentences. I remember that the smoothness estimates from the residuals of the fMRI data, which could be also found in the second paper "The Past﹣ 3dClustSim and “The Bug.” The first problem was particular to AFNI: there was a bug in 3dClustSim. This program works by generating a 3D grid of N(0,1) iid random deviates, then smoothing them to the level estimated from the residuals of the FMRI data model, then carrying out voxelwise thresholding followed by clustering to determine the rate at which clusters of different sizes occur at the various voxelwise thresholds."

If I just want to perform group comparison of ReHo/ALFF between two groups, there is actually no first-level analysis in my experiment, so how can I estimate smoothness from first-level analysis? Or estimating smoothness from the first-level analysis is just used in the task-fMRI data preprocessing?

Any help would be greatly appreciate.

YAN Chao-Gan

Tue, 08/30/2016 - 01:18

Hi Feng,

You should estimate the smoothness based on 4D residuals, according to the manual of 3dFWHMx.

If you use DPABI perform two-sample t-test, please edit y_TTest2_Image.m

Line 79


[b_OLS_brain, t_OLS_brain, TTest2_T, r_OLS_brain, Header] = y_GroupAnalysis_Image(DependentVolume,Regressors,OutputName,MaskFile,CovariateVolume,Contrast,'T',0,Header); 


[b_OLS_brain, t_OLS_brain, TTest2_T, r_OLS_brain, Header] = y_GroupAnalysis_Image(DependentVolume,Regressors,OutputName,MaskFile,CovariateVolume,Contrast,'T',1,Header); 


Then you will get the 4D residual to estimate the smoothness.