During our postprocessing we concatonate 4 runs via 3dTcat, and subsiquently input this file into 3dDeconvolve.

We have been using the "-rlt" option for the 3dTcat program because we assumed we should remove the linear trend prior to deconvolution.
We also have been using the "-polort 1" option in 3dDeconvolve to remove the baseline and the baseline drift.

Recently we found that the output coefficients are a little different when you use the "-rlt" option (in 3dTcat) and when you don't use it, for data that is eventually deconvolved with the "-polort 1" option.

Are there negative affects on the data by, theoretically, removing the baseline and linear trend twice?

We know the two options are doing two different things. One involing the +time data itself and the other involving the model calculated by multiple regression, but we still are not sure of the implications of both or either.

The most pragmatic question though, is whether we should rerun our previously "-rlt"ed data, continue to be consistant(aka run 3dTcat with "-rlt" and 3dDeconvolve with "-polort 1"), or just stop running 3dTcat with the "-rlt" option?

Any input would be appreciated.
Thank you so much for your time!

Jeremy