Thanks Jaakko! That didn't quite work, but was close enough to help me
figure it out. For the record, the working solution is:
# set up dummy first trial to set array dimensions
events_adj = [[0,0,0,]]
for key, value in event_id.items():
events_tmp, lag_tmp = mne.event.define_target_events(events,
sound_onset, value,
raw.info['sfreq'],
code_tmin, code_tmax,
new_id=event_id_adj[key],
fill_na=None)
events_adj = numpy.concatenate((events_adj, events_tmp))
The trick was initialising the events_adj array with a [0, 0, 0] entry,
otherwise the np.concatenate call fails. This doesn't strike me as the most
elegant solution, but it does work!
Also, you were correct that there's no need to sort the new events array
afterwards; MNE manages just fine.
Best,
Aaron