I am confused by the warning in the documentation for 3dTshift:
"Please recall the phenomenon of 'aliasing': frequencies above 1/(2*TR) can't be properly interpolated. For most 3D FMRI data, this means that cardiac and respiratory effects will not be treated properly by this program."
What if I am planning to correct for physiological noise using RETROICOR? This correction should be done by regression on a slice by slice basis, and it will remove noise from the time course of the data. In this case, would regression of the physiological noise prior to 3dTshift be advised in order to perform temporal interpolation on cleaner data?