- Do not mix non-seasonal and seasonal series in one input file, and do not to use TD_AUTOARIMA on mixed multi-series input.
- Do not mix seasonal data with different period attributes in one input file, and do not to use TD_AUTOARIMA on such mixed multi-series input.
- For seasonal input series, you must input correct period in the functional parameters. TD_AUTOARIMA cannot determine the period from the input data. For example, if your input data is monthly data and you do not include the period in the functional parameter, then TD_AUTOARIMA uses the default period value (1) as non-seasonal data. Incorrect periods result in inaccurate best-model suggestion in the TD_AUTOARIMA result.
- Expect long run times when input data is entered without order constraints, especially for a long seasonal period or large time series.
- Maximum value of p+q+P+Q is 5 when model selection is not stepwise.
- TD_AUTOARIMA is based on the best-information criterion selection process. The returned model is the best model and does not need more validation like TD_ARIMAVALIDATE to compare the AIC and BIC with different model candidates. If you run TD_ARIMAVALIDATE on an ART from TD_AUTOARIMA, an error message is displayed.
- TD_ARIMAFORECAST accepts ART result from both TD_AUTOARIMA and TD_ARIMAESTIMATE. After using TD_AUTOARIMA, you do a time series forecast using one of the following:
- Run TD_ARIMAFORECAST directly against ART result generated from TD_AUTOARIMA.
- Extract the ARTICANDORDER layer to get model order. Next, run TD_ARIMAESTIMATE using the model order, and then run TD_ARIMAFORECAST on the ART result from TD_ARIMAESTIMATE.