I performed a time series forecast using auto_arima from the pmdarima package. I know that this package is based on the statsmodel SARIMAX package.
Using the command: fit.predict_in_sample(ARIMA_input, dynamic=None), gives results that are scaled/normalized around 0. I would like to transform the in sample predictions back this to plot it with my input data. Does anyone know how it is transformed? I searched in the source code from pmdarima but I cannot find anything. Using statsmodel SARIMAX the in sample predictions came in the same scaling as my input.
Note: my data is not seasonal so I just use an ARIMA with the statsmodel SARIMAX.
Moreover, if I use the order given by the auto_arima fit from pmdarima and use it with statsmodel SARIMAX, I get different forecast results (pmdarima's forecast is very plausible and SARIMAX's is just a straight line). It seems I am not seeing what's done different. Maybe someone of you understands it better and can help me?
If you need more information, I am happy to provide it.
