Hello malf:
I am confused by your question. The "-rlt++" option of 3dTcat will add the
overall mean back in, so the "b0" coefficient should NOT be zero. Are you
sure that "it removed 'b0+b1t' together"? What are typical values for "b0"
contained in the output bucket dataset?
As Bob said, it is not necessary to use the "-rlt" option of 3dTcat, since
3dDeconvolve will fit a separate constant plus linear trend to each of the
concatenated runs. See the description of the "-concat" option in Sections
1.2.12 and 1.4.6 of the documentation in file 3dDeconvolve.ps.
Since you are using the baseline coefficient to standardize the data, it is
necessary for 3dDeconvolve to write the baseline coefficients to the output
bucket dataset. Hence, the command line should NOT include the "-nobout"
or "-nocout" options.
Note that, for concatenated runs, it may be preferable to "standardize" the
time series data prior to deconvolution. For a discussion of this alternative
approach, see the messages at or about:
[
afni.nimh.nih.gov]
Doug Ward