First off, thank you for developing such an amazing library—StatsForecast has been enormously helpful in my projects! I’m especially interested in the in‑sample generator method forecast_fitted_values().
While trying to obtain multi‑step in‑sample predictions from an ARIMA model, I called: model.forecast_fitted_values() . I was wondering if fitted_values can be produced with h > 1 , but tracing the code I noticed there is a restrintion in this lines of code. Is there a theoretical / statistical reason to forbid h > 1, or is multi‑step support simply not implemented yet? Any recommended workaround to generate multi‑step in‑sample forecasts for models?
First off, thank you for developing such an amazing library—StatsForecast has been enormously helpful in my projects! I’m especially interested in the in‑sample generator method forecast_fitted_values().
While trying to obtain multi‑step in‑sample predictions from an ARIMA model, I called: model.forecast_fitted_values() . I was wondering if fitted_values can be produced with h > 1 , but tracing the code I noticed there is a restrintion in this lines of code. Is there a theoretical / statistical reason to forbid h > 1, or is multi‑step support simply not implemented yet? Any recommended workaround to generate multi‑step in‑sample forecasts for models?