pymc-experimental
pymc-experimental copied to clipboard
Allow forecasting with exogenous variables
Enable scenario-based forcasting in VARMAX, SARIMAX, and structural models with regressions.
Couple to-dos before its ready for review:
- [x] Carefully check the timing on the forecasting. Things seem to be shifted forward by 1 or 2 periods.
- [ ] Recycle the exogenous forecasting code to allow in-sample counterfactual scenarios using
sample_conditional_posterior
orsample_conditional_prior
. - [ ] Update relevant example notebooks to show how to use the scenario tools
Otherwise things seem to be working