Thanks, @paulmccarthy! I’ll try to translate this into choices in FEAT GUI. Please correct me if I’m getting it wrong.
1- Pre-filtering: in FSL FEAT GUI for level1 GLM, it is done by turning on temporal highpass filtering in the Pre-stats tab (which adds the fslmaths -bptf command) . If I choose this option, I will then have to “apply temporal filtering” to my EVs during full model setup.
2-Modeling low frequency drifts: in fmriprep, it is done through calculation of cosine-basis regressors. To apply these in FSL FEAT, I will have to include these cosine regressors as “confound EVs” in the stats tab. If I choose this option, I should not turn on temporal highpass filtering in FEAT pre-stats, and I should not “apply temporal filtering” to my EVs during full model setup.
Regardless of the above choice, I can safely “add temporal derivative” in my full model setup.