11-18-2011 04:34 PM
The 'selection diagnostics' of SAS Forecast Sever analyzes each series for intermittency, trend, seasonality, and other properties based on a battery of statistical tests..
If a model's components do not match these properties, it is rejected.
For example, if a series is determined to be non-seasonal and the model contains seasonal components, it is rejected.
Rejection prevents over-parameterization; otherwise, seasonal models would always win in many cases.
Some additional information about why a model was rejected specifically can be found in the OUTSUM and OUTSTATSELECT data sets of HPFENGINE - see: