Thanks Rick, I followed your advice and things seem to be working very well.
Our input data are not scaled, and as a result, output beta coefficients are pretty big, somewhere around 200-500 (using "-basis_normal 1"). I'd like to scale, or mean normalize, the input data before running 3dDeconvolve. Specifically, I'd like to set the data so that values are scaled such that their mean is 100, and every value represents the variation from the mean as a percent... converted into % signal change.
Is there command(s) to do this operation?
Thanks a lot!
Akiko