Hi all,
I am running a first and second-level analysis with nilearn on fmriprep-preprocessed data (and using nuisance regressors from the confounds file), My z-score/statistical map outputs look plausible, however my beta maps look like they just contain noise/artefact (either midline activations or single ‘outlier’ voxels in random locations). I am trying to figure out what is going on and is it possible that it could just be a scaling issue of my task regressors (compared to all the others)?
This is what part of the design matrix looks like with the first 4 columns being the task events and the last couple of columns removal of outlier volumes based on FD.
Thanks all for your help!