Paired t-test results looking off

AFNI Message Board
1

Hello,

I was wondering if someone can please assist with viewing the results of a paired t-test. The results look off in the sense that connectivity shows in regions like the ventricles, cerebellum, and even outside the brain. Is this correct?

Please see the screenshots of the results attached (includes paired t-test and paired t-test with clustsim).

Please see the script I am running below:

cd /Volumes/RJ/INSAR/groupanalyses/
3dttest++ -prefix rFG_ASDvsTD1.nii.gz -AminusB
-setA
/Volumes/RJ/INSAR/post/ASD148E_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD167C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD168C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD171C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD181C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD184C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD192C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD194C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD196C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD200C_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD204_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD205B_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD207B_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD300_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD301_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD302_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD303_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD304_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD305_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD306_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD307_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD308_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD4004_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/ASD4005_FC_rFG_R-Z.nii.gz
-setB
/Volumes/RJ/INSAR/post/42003_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42004_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42007_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42008_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42009_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42010_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42013_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42015_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42016_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42019_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42022_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42023_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42027_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42029_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42030_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42031_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42033_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42035_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42036_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42038_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42039_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42040_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42041_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/42042_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/TD106E_FC_rFG_R-Z.nii.gz
/Volumes/RJ/INSAR/post/TD163C_FC_rFG_R-Z.nii.gz

end

I’ve tried the above script with “” before and after the data file names but getting the same result. Any advice or suggestions you can provide would be greatly appreciated!

Thanks so much!
RJ

paired t-test.png
paired t-test.png1826×1476 618 KB

paired t-test with clustsim.jpg
paired t-test with clustsim.jpg1826×1476 628 KB

2

RJ,

the results of a paired t-test

I assume that you’re comparing two groups. So, it would be a conventional two-sample (not paired) t-test.

The results look off in the sense that connectivity shows in regions like the ventricles, cerebellum, and even outside the brain. Is this correct?

The results in some of those regions you mentioned are likely random noise, especially when sample sizes are not large.

3

Hi Gang,

Thank you for your reply. Yes, that is correct, comparing two groups - thank you for clarifying.

So just to confirm, the random noise showing in the ventricles, cerebellum and outside the brain is fine in this case?

Thank you!
RJ

4

the random noise showing in the ventricles, cerebellum and outside the brain is fine in this case?

The model at the population level in this case (two-sample t-test) is pretty simple and straightforward. By “random noise”, they could include issues such as data quality (e.g., alignment), processing, head motion, physiological confounds, etc.