GLM Analysis with multiple files (3dTcat or 3ddeconvolve?)

AFNI Message Board
1

Hello,

While I was building my GLM, I was unsure which would be the best approach. I have 2 localizer tasks in my study.

In order to build my ROI masks,
I was planning on concatenating the data by using 3dTcat before I put it into my GLM.

  1. Is it better for 3ddeconvolve to handle my concatenation or may I do it manually?
  2. My current timing files look something like this, is this the most optimal format when 3ddeconvolve will concatenate my files?
32.1	64.3	87.5	96.4	112.5	164.2	178.5	294.5	308.8	337.4	348.1	394.5	414.1	428.4	437.3	455.2	515.9	562.3	598	623	
1.8	19.6	94.6	121.4	135.7	153.5	160.7	174.9	210.6	291	296.3	342.7	399.8	407	435.5	519.4	580.1	605.1	621.2	637.2

Thank you very much.

2

Hello,

Yes, while either way will work, it seems slightly preferable to pass one dataset per run. That tells 3dDeconvolve where the run breaks occur. Otherwise, you will need a -concat option, not a big deal.

Either way, that timing file looks good.

-rick

1 Like
3

That makes sense. Thank you. I do have another question about the GLM script.

My current GLM input dataset is in MNI space and the GLM results gets automatically saved as +orig file format.

Is it alright to use 3drefit to convert this file into tlrc or MNI space so I can overlap it with my structural scans (also MNI space)? I am trying to see the activation clusters for my ROI mask, so I would like to see it nicely overlapped. Below is the code I use.

3drefit -space MNI -view tlrc \
    ${OUT_DIR}/results+orig

3drefit -space MNI -view tlrc \
    ${OUT_DIR}residuals+orig
4

Yes, that is certainly fine. I suspect your input is in NIFTI format with an sform_code of 2, which is an ambiguous code. Is that the case? Check with:

nifti_tool -disp_hdr -field sform_code -field qform_code -infiles INPUT.nii.gz

If that is the case, you will indeed want to fix it at some point: how such files are written in the first place, after the fact, or after the regression. Note that the code for MNI is 4 and the code for a generic standard space (since there are many) is 5.

-rick

1 Like