TD_SIGNIF_RESIDMEAN

The Significance of Residual Mean statistical test is aimed at assisting the data scientist in selecting the model candidate which is the best choice for the data they are modeling. The test which works off the residuals generated during the model validation / model selection processing phase. As mentioned previously, typically the data scientists divide their series sample set into two portions: the first portion to be used for the “model fitting” exercise; and, the second portion to be used for the “model validation” exercise. The TD_SIGNIF_RESIDMEAN function is usually executed against the residuals produced during the second “model validation” exercise, meaning against the “in-sample” forecasted data points.

Input ports

1. Type: DB Session
   Teradata Connection
   Connection to a Teradata Database Instance
2. Type: Table
   Input
   This function can accept either a single univariate series (containing the in-sample generated residuals) as an input, or alternatively can accept an ART table containing the residual results from a previously run validation (TD_ARIMAVALIDATE) operation. For the case in which the input is a univariate series, the payload should contain a reference to a single field containing the residuals: the single field references the calculated residual: original_value minus calculated_valueFor the scenario in which the residuals reside within the tertiary result set associated with an ART table, then the caller merely needs to pass in the TABLE_NAME identify the ART containing the residuals - the DBS software will perform the extraction of the original series, calculated series, and residuals as part of the execution of this function.

Output ports

1. Type: Table
   output of TD_SIGNIF_RESIDMEAN
   output of TD_SIGNIF_RESIDMEAN