Hi Kristin,

The GAM basis function is intended to be for a
short event (around 1 second, or less). Since
your events are all at least 3 seconds, GAM is
not really accurate enough.

One basic question is whether you want to actually
compare the betas across classes (e.g. contrast
cue vs. anticipatin or stimulation).

Assuming so, consider:
cue: BLOCK(3)
anticipation: AM1 and dmBLOCK
stimulation: BLOCK(6)

Note that using BLOCK(3,1) or BLOCK(6,1) would
make the comparison across classes biased toward
the longer events (stimulation).

To apply this in afni_proc.py, consider:

-regress_stim_times cue_times.txt ant_times.txt stim_times.txt \
-regress_stim_labels cue ant stim      \
-regress_stim_types times AM1 times    \
-regrsss_basis_many 'BLOCK(3)' dmBLOCK \

- rick