default value for 3dTProject bandpass

AFNI Message Board
1

Hello!

Is there a default value for the parameter -bandpass in 3dTProject?

I ran the commands

3dTproject -input bold_orig.nii.gz -polort 1 -prefix bold_project_1.nii.gz

3dTproject -input bold_orig.nii.gz -polort 1 -prefix bold_project_2.nii.gz -bandpass 0 1

I expected bold_project_1 and bold_project_2 to have exactly the same voxel values, but they don’t, so I’m wondering if the command above without -bandpass uses a range different than [0 1].

Thank you!

Best regards,
Sam

2

Hi, Sam-

When I ran those commands (use the AFNI_data6/afni/epi*HEAD dset, in the Bootcamp data), running:


3dTproject -input epi_r1+orig. -prefix AAA -polort 1 -verb
3dTproject -input epi_r1+orig. -prefix BBB -polort 1 -bandpass 0 1 -verb

… I used “-verb” to output the regressor files.

First, I noticed a bit of difference in the output terminal text:
For AAA:


++ Setting up regressors
++ 1 Blocks * 2 polynomials -- 2 polort regressors
++ 152 retained time points MINUS 2 regressors ==> 150 D.O.F. left

For BBB:


++ setting up stopband frequency mask
 + Block #0: 152 time points -- 1 stopband regressors
++ 1 Blocks * 2 polynomials -- 2 polort regressors
++ 152 retained time points MINUS 3 regressors ==> 149 D.O.F. left

So, there is an extra regressor in the BBB case, because of the bandpassing.

The output AAA.ort.1D file had 2 cols, one for each regressor, with the top few lines looking like this (first col is the flat/baseline regressor from polor=0, and second is the polort=1 linear one):


              1             -1
              1      -0.986755
              1       -0.97351
              1      -0.960265
              1       -0.94702
              1      -0.933775
              1       -0.92053

In BBB.ort.1D, those same polort columns exist, as well as a new column, from the bandpassing:


              1             -1              1
              1      -0.986755             -1
              1       -0.97351              1
              1      -0.960265             -1
              1       -0.94702              1
              1      -0.933775             -1
              1       -0.92053              1

This last column represents the highest frequency possible to capture in the data, oscillating at each measured timepoint. It is typically a “low power” frequency in practical data, and indeed the difference between the AAA and BBB output dsets is generally tiny.

It seems like when including the “-passband …” option, this highest frequency is included in the stopband. And actually, the lowest frequency in your stopband is also included, but it just to happens to be the “flat” baseline, which is part of your polort, so you don’t see it. You would see it explicitly if you turned off the polorts:


3dTproject -input epi_r1+orig. -prefix CCC -polort -1 -bandpass 0 1 -verb

There are 2 degrees of freedom removed, and CCC.ort.1D has 2 columns: the freq=0, and the max frequency.

–pt