Dear CONN Community,

I am unclear about what data is output and what spatial unit is used for linear detrending is applied in the CONN preprocessing steps. I understand that this method of cleaning the signal involves regressing out the linear term to remove the potential influence of signal drift.

Question 1: Is the linear (or other) detrending applied at the voxel time-series level or the averaged ROI time-series level?

Question 2: When "regressing out" the linear term, are the residuals (observed-predicted) from this model output and passed on to the second-level analysis, or are the predicted values passed on for second-level analysis? 

Thank you for your help!

Best Wishes,
Allison