CODEWITHSUNDEEP
pythonstatsmodelspmdarima
siebenkaese
siebenkaese

Reputation: 57

How to update/append model estimated via pmdarima without restimating the parameters

I am using the pmdarima package fit a model with an exogeneous variable:

    processor = preprocessing.LogEndogTransformer()
pipe_log_onw_price_model_2022 = pipeline.Pipeline([
    ("log", processor),
    ("arima", pmd.arima.AutoARIMA(stepwise=True, 
                                  trace=True,
                                  start_p= 1,
                                  m=12,
                                  suppress_warnings=True))
])
pipe_log_onw_price_model_2022.fit(train_2022[["onw_price"]],X=train_2022[["onw_pen"]])

I want to use the model for a simulation employing custom errors/innovations and without reestimating the model parameter (they should stay as they are). The pmdarima package only includes the function "model.update" which updates the parameters of the model. For my purposes the statsmodel method statsmodels.tsa.arima.model.ARIMAResults.append seems to be appropriate.

Is there an easy way to use something like the append function on the pipeline estimated in pmdarima?

I tried directly accessing the statsmodel in the pipeline and calling append on this but than I would need to apply the preprocessing step in the pipeline again, which I would want to avoid.

Additionally, I tried pipe_log_onw_price_model_2022.update(update_y, update_x, arima__update_params=False), but this still updates the parameters.

Thanks!

Upvotes: 0

Views: 78

Answers (0)

Related Questions