Hello,
I have some questions regarding the new option -acf both for 3dFWHMx and 3dClustSim. To estimate the blurring across all subjects to determine acceptable cluster size at .005, I ran 3dFWHMx on each subject’s residuals (errts+tlrc). I averaged the three parameters across subjects to get the mean radius (r) and two fitted parameters (a & b). I entered these parameters into the 3dClustSim using the acf option fully expecting more conservative cluster size values but instead acceptable cluster were more liberal. When I estimated blur size using the old method acceptable cluster size was 71 voxels but with the new method (acf) it was 44 voxels.
1) Since acf is supposed to make clusters more conservative, is it possible to see this kind of decrease in acceptable cluster size when switching between methods for estimating the blur?
Looking into this further, a colleague noticed that when running afni_proc when 3dFWHMx is called the data is detrended (option –detrend). Since when I ran 3dFWHMx originally, I did not use the detrend option we thought this might be why cluster size became more liberal. So we reran 3dFWHMx on each subject using the same command adding the detrend option. The only other difference is that we fed 3dFWHMx the list of uncensored volumes. Results here showed a more conservative cluster size of 104 voxels.
2) Does it make senses that the difference in cluster size is due to detrending when calling 3dFWHMx? If so, would it be wise to make sure this option is called when estimating effective blur when getting the parameters for the –acf method?
Thanks for your help
Steve
