Hi Bob,
In part based on Molly Bright's recent paper, I'm following up on removing/mitigating autocorrelation in our resting-state data at the nuisance regression step. Following up on your previous post, wouldn't I want to use -Rwherr as the output of 3dREMLfit for subsequent connectivity analyses? My understanding was that -Rwherr keeps the pre-whitened residuals, whereas -Rerrts just subtracts the fitted regressors without handling the temporal autocorrelation. I wanted to check with you on this to make sure I wasn't overlooking something conceptually important that led you to recommend -Rerrts here.
Thanks much,
Michael