Dear Daneil,

Thanks for your suggestions!

> 1. In your 3dcalc command, -datum short or float would be
> better. Byte only gives you very limited resolution.

I havd changed the command into
3dcalc -datum short -a3 Subject_BF_First+tlrc -b SubjectROI_F_Amy+tlrc -prefix SubjectAmy_BF_First -expr "step(abs(a)-6.674)*b"

> 2. Also I assume the -a3 is the fourth sub-brick in the
> zero-based indexing.

Yes, my typo. The fourth brick is what I refer to.

> 3. You need quotes of some kind around the dataset name that
> specifies a sub-brick in the 3dROIstats command.

I have modified the command like this
3dROIstats -nzvoxels -mask SubjectAmy_BF_First+tlrc 'Subject_BF_First+tlrc[0]' > Subject_Amy_BF_First.txt

> 4. The number of voxels in the mask is very low for all you've
> shown. Have you looked at the mask in the AFNI viewer?
> 3dBrickStat (formerly 3dMax) can give a count too to get the
> number of voxels in your mask. You can use that or 3dmaskave to
> get the means with masks too.

The number of voxes in the mask is around 250. I reckon that it is reasonable because I am looking at the amygdala only.

I reran the new script but still got weired results in some participants. Will try the new 3dROIstats tomorrow.

Cheers! Shen-Mou