I have seen in several papers that grand mean scaling is a frequent step included before the first level analysis. I don’t know if this step is mandatory and when it should be included. In brief, my preprocessing pipeline includes the following steps: (i) motion corregistration, (ii) retroicor, (iii) slice timing correction, (iv) despiking, (v) normalization to the MNI template, (vI) nuisance regression, and (vII) smoothing.
Should I perform the grand mean scaling just after smoothing?
Different softwares recommend/promote scaling in different ways. Each carries different implicit assumptions. Grand mean scaling is one particular manner for scaling data; it is not one we recommend. The FMRI scaling method implemented in AFNI’s afni_proc.py–what happens when you use the “scale” block–translates the arbitrary units of FMRI datasets into time series that are interpretable as local BOLD percent signal change; each time series is scaled by the “baseline” (~mean) value of a given voxel. This scaling seems to be appropriate for the underlying mechanisms of BOLD-modulated signals, as well as for allowing comparisons of effect estimates (AKA “beta” values or coefficients) across a brain and across a group.
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.