Hi everybody,
I am trying to do an ARIMA forecast of hospital examinations with daily time-series data including 1000+ observations.
The current fit is a SARIMA(1,1,1,7) model in order to mitigate the seasonality in the data.
My problem is that my residuals are non-normal after fitting the correct ARIMA model. I have log-transformed and differenced the primary variable.
The qqplot of the residuals vs. the main variable can be seen here:
I am unsure of what to do next with the model, as I suspect the forecasts will be invalid.
Should I consider fitting the model with differently-distributed error terms?
If so, how should I do that?
Thank you,
Martin
I am trying to do an ARIMA forecast of hospital examinations with daily time-series data including 1000+ observations.
The current fit is a SARIMA(1,1,1,7) model in order to mitigate the seasonality in the data.
My problem is that my residuals are non-normal after fitting the correct ARIMA model. I have log-transformed and differenced the primary variable.
The qqplot of the residuals vs. the main variable can be seen here:
I am unsure of what to do next with the model, as I suspect the forecasts will be invalid.
Should I consider fitting the model with differently-distributed error terms?
If so, how should I do that?
Thank you,
Martin
