3dISC Number of Rows and Possible Pairs

AFNI Message Board
1

AFNI version info (afni -ver): Precompiled binary linux_rocky_8: Apr 10 2026 (Version AFNI_26.1.00 'Balbinus')

The current setup for our experiment is to have participants perform two trials (one auditory and one visual) of the same length. We are wanting to look at the differences of a specific regressor between the visual and auditory trials. To create the data table, I have run 3dt_correlate for each type of trial separately as we do not want to directly compare the audio vs. visual. This means we have unique participant codes in the data table for both the visual and auditory trial. With the call below we are receiving the following output:

"Read 1099 items
Warning message:
In process.ISC.opts(lop, verb = lop$verb) : NAs introduced by coercion

++++++++++++++++++++++++++++++++++++++++++++++++++++
***** Summary information of data structure *****
12 subjects in set1 : SPNF_004 SPNF_005 SPNF_006 SPNF_007 SPNF_008 SPNF_009 SPNF_011 SPNF_012 SPNF_013 SPNF_015 SPNF_018 SPNF_019
12 subjects in set2 : SPNF_005 SPNF_006 SPNF_007 SPNF_008 SPNF_009 SPNF_011 SPNF_012 SPNF_013 SPNF_015 SPNF_018 SPNF_019 SPNF_020
13 subjects in total: SPNF_004 SPNF_005 SPNF_006 SPNF_007 SPNF_008 SPNF_009 SPNF_011 SPNF_012 SPNF_013 SPNF_015 SPNF_018 SPNF_019 SPNF_020
156 response values

** Error:
Error: the number of rows/files, 156, is not equal to 78 - the
possible subject pairs!"

This error makes sense as we do have double the number of possible pairs since we have two trials. We are wondering if we can move forward without having to specific the unique participant codes such as adding '_aud' or '_vis' to the end of our participant IDs. We are hoping to capture the differences in modality by our regressor of interest through the 'is_vis_x_aq' contrast.

3dISC -prefix ${prefix} -jobs 4 -r2z \
        -mask MNI_mask_epiRes.nii\
        -model 'sum_aq+Sum_motion+is_vis+is_vis_aq+(1|Subj1)+(1|Subj2)'\
        -qVars 'is_vis_aq,is_vis,Sum_motion'\
        -gltCode ave '1 0 0 0 0' \
        -gltCode is_vis_x_aq '0 0 0 0 1'\
        -gltCode sum_aq '0 1 0 0 0'\
        -gltCode motion '0 0 1 0 0'\
        -gltCode mo_aq '0 0.5 0.5 0 0'\
        -gltCode mo_is_vis '0 0 0.5 0.5 0'\
        -gltCode sum_aq_vis '0 0.5 0 0.5 0'\
        -dataTable @${data_table}
2

A few quick questions to make sure I understand the setup:

  • Does each input file represent the correlation between two participants under the visual or auditory condition?

  • What do the various variables in your model formula stand for?

  • What specific scientific questions are you trying to address with this model?

Gang Chen

3

Hi Gang!

Each input file represents two participants under either the visual or auditory condition. There are no files that represent the correlation between one visual and one auditory trial.

I have set up a new model that I put into a code block below that has an updated version of the contrasts and variables. This model still yields the same error as described originally. is_vis represents a variable that is a 1 if the trial is visual and a 0 if the trial is auditory. sum_aq represents a variable that is the sum of a behavioral score for autistic traits. sum_motion is actually an averaged motion from motion_enorm.1d from afni_proc.py for the participants for the trial. is_vis_aq is the is_vis variable multiplied by the sum_aq variable. motion_is_vis is the sum_motion variable multiplied by the is_vis variable.

research question 1: how is AQ associated with auditory stories (should be answered by the aud gltCode below)

research question 2: how is AQ associated with visual stories (should be answered by the vis gltCode below)

research question 3: how is AQ associated with stories independent of modality (should be answered by the sum_aq gltCode below)

research question 4: how are the differences in AQ associated with the differences in ISC between visual and auditory stories (should be answered by sum_aq_vis gltCode below)

Thank you so much!

Caitlin

3dISC -prefix ${prefix} -jobs 4 -r2z \
        -mask MNI_mask_epiRes.nii\
        -model 'sum_aq+Sum_motion+is_vis+is_vis_aq+motion_is_vis(1|Subj1)+(1|Subj2)'\
        -qVars 'is_vis_aq,is_vis,Sum_motion'\
        -gltCode ave '1 0 0 0 0 0' \
        -gltCode vis '0 0 0 0 1 0'\
        -gltCode sum_aq '0 1 0 0 0 0'\
        -gltCode motion '0 0 1 0 0 0'\
        -gltCode mo_is_vis '0 0 0 0 0 1'\
        -gltCode sum_aq_vis '0 0.5 0 0 -1 0'\
        -gltCode aud '0 1 0 0 -1 0'\
        -dataTable @${data_table}
4

Hi Caitlin,

Because 3dISC was not implemented to handle within-individual variables, I recommend addressing each of your four research questions through four separate analyses.

I assume all input files are Fisher-transformed Z-scores. If they are raw correlation coefficients, convert them to Z-scores prior to running 3dISC.

Addressing your questions specifically:

research question 1: how is AQ associated with auditory stories (should be answered by the aud gltCode below)

Use only the input files corresponding to auditory stories.

research question 2: how is AQ associated with visual stories (should be answered by the vis gltCode below)

Use only the input files corresponding to visual stories.

research question 3: how is AQ associated with stories independent of modality (should be answered by the sum_aq gltCode below)

For each participant pair, calculate the average Z-score across auditory and visual stories, and use those pair-wise averages as the input dataset.

research question 4: how are the differences in AQ associated with the differences in ISC between visual and auditory stories (should be answered by sum_aq_vis gltCode below)

For each participant pair, calculate the difference in Z-scores between the two modalities, and use those differences as the input.

Let us know if this strategy achieves what you are looking for.

Gang Chen