Repository navigation
[BUG]: Result from pyciemss.sample do not reflect interventions accurately. #624
Description
When we use pyciemss.sample to simulate a model with interventions, the output is only showing the model state vector as if only interventions are considered when they occur within [t_previous, t_current) for a given t_current.
The state/parameter data in the output definitely should reflect intervened values at the time of the interventions. This is particularly relevant in the case of dynamic interventions.
Activity
liunelson commentedon Nov 26, 2024
@SamWitty
I don't recall whether a fix for this issue has been merged already - I recall that you offered to add an extra function evaluation at the timepoint of the intervention and we agreed on it.
djinnome commentedon Dec 16, 2024
Hi @liunelson is this a consequence of making
logging_step_size=1? See if you still get unexpected behavior iflogging_step_size=0.1.liunelson commentedon Feb 27, 2025
@djinnome
This is the issue that we discussed earlier today. A solution would be to include for every intervention sample point at time
t = t_interv + eps, eps << 1.