Hi Alfonso,

My QC-FC median motion plot matches the null hypothesis 94.3% of the time (using a conservative threshold, quadratic effects, and second-order derivatives for the realignment parameters).

Thus, I should exclude some subjects from my analyses. So far, I have identified two extreme outliers: one in norm_str and one in PVS. Excluding these two subjects increased the match to 94.8%.

I am wondering which parameter I should prioritize when selecting additional subjects for exclusion (PVS, DOF, mean motion, norm_str, norm_func, BOLDstd, or gcor). Most participants appear to be mild outliers in only one of these variables, and removing those with the most deviant values (one at the time) does not significantly improve my match. This suggests that I would need to exclude more than one additional subject to reach 95%.

However, I noticed that if I exclude one subject who is a moderate outlier in at least two variables (mean motion, where they are the second most deviant outlier, and norm_str, where they are the third most deviant outlier), my match increases to 95.3%, while only removing three subjects in total.

I understand that I should aim to remove as few participants as possible (I'll run an MVPA), but I am unsure whether my justification is valid or if I should prioritize other variables (e.g., PVS, BOLDstd, or Gcor).

I’d appreciate your thoughts on this!