Thanks Daniel,

I've worked out the RAI specific issue. I've make sure that the 3dmaskdump to 3dUndump pipeline is consistently in RAI, which I then 3dresample to LPI to match the original data orientation. When putting together this analysis pipeline I worked with the ijk system but I found keeping things in RAI was easier.

Moreover, the group data set is aligned with the HaskinsPeds_NL_template (that I have down sampled to 2.5mm). See attached.

Breaking down the commands:

3dUndump -xyz -datum float \
-prefix /Volumes/NBL_Projects/NBL_WorkingPapers/HomeMath_RSA/groupanalyses/ROI/Group_within_ratio_Region_digitratio_t_RAI \
-master /Volumes/NBL_Projects/NBL_WorkingPapers/HomeMath_RSA/ref/atlas/BN_Atlas_246_2.5mm_to_HaskinsPediatric_RAI+tlrc.HEAD \ /Volumes/NBL_Projects/NBL_WorkingPapers/HomeMath_RSA/groupanalyses/ROI/Group_within_ratio_Region_digitratio_t_voxel_data.txt

Group_within_ratio_Region_digitratio_t_voxel_data.txt

This is the output from a RSA analysis where I have assigned the tvalue of the model fit for each ROI to all the voxels in that ROI. I've done this by editing the output of the 3dmaskdump of the atlas of ROIs (also -master here). I've ensured the output is written in RAI orientation since the this is the orientation output by 3dmaskdump and 3dUndump

I've emailed a zip file to both Daniel and Peter with the the input to this command as well as the function I use to create the txt and *BRIK/HEAD files that are misaligned to the HaskinsPeds freesurfer output.