Teo,
I am not so sure what your purpose of these step is. Is it preparation for group analysis, or is it simply for converting individual result into standardized format?
The explanation about performing smoothing after running 3dDeconvolve is that we wanted to keep the signal as close as possible to the raw signal.
It seems most people tend to spatially smooth their data before they run regression analysis. I guess you can argue the other way around.
Is it mandatory to use as func: beta coefficients and as threshold : t stats or partial F stats?
In fact last time I was confused whether the subbrik you extracted was regression coefficient or its t value. If you want to do group analysis based on your smoothed t values, this might not be a good idea because most statistical analyses including ANOVA are supposed to be done on raw data such as regression coefficients with assumption of normal distribution.
However, if you just want to graphically demonstrate individual result, it is up to you to select the Func dataset. The thrshold dataset has to be some statistical values, and the threshold slider on AFNI GUI decides the cutoff value. Although most poeple tend to use regression coefficient as Func dataset so that the color-coded patterns in the brain are interpreted as percent signal change, it is not mandatory at all and theoretically you can color code any subbrik as you want.
Gang
viewing