Description
Currently, StatsForecast models provide point forecasts and prediction intervals, but lack a direct method to generate predictive samples from the fitted models.
A built-in method to generate predictive samples from fitted models such as generate would make StatsForecast more competitive with R's fable ecosystem. This would also allow us to switch from fable to nixtla in our workflow which relies on using sample trajectories.
Use case
A method to generate predictive samples/full predictive distribution from fitted models.