Hi,

In one dataset that I use in most of the subjects the histogram before removing confounds is drastically shifted (example attached). Based on what is written in the FAQ and the common sense the major source for such a shift is movements in the scanner. However, there are some case when the movements were moderate (see bottom panel), but the first dimension (green) of the white matter shows huge variability and the histogram is shifted. So, movements seem to be not only reason for histogram shift. More generally, how can I trust the results after the shift is fixed? From my experience with fMRI GLM analysis serious movements can completely abolish the effect. Interestingly, in this specific dataset with histogram shift (and many subjects with movements) the effect that I get is indeed incompatible to what I get in sets without histogram shift.

Thanks for help,
Vadim