TD_SIGNIF_RESIDMEAN is a statistical test to determine if the passed-in residual series can be classified as being white noise. White noise is a series that has zero mean, no evidence of serial correlation, and homoscedastic variance. It is a random sequence of independent values that are identically distributed. The function uses the residuals generated during model validation of the forecasted data points. The expectation is that if the appropriate ARIMA model has been chosen, then any leftover residuals are white noise.
The following procedure is an example of how to use TD_SIGNIF_RESIDMEAN:
- Divide the sample data into two sets. One set is used to fit the data to the model, and the other set is used to select between models.
- Use TD_ARIMAESTIMATE to determine the coefficients associated with the ARIMA model.
- Use TD_ARIMAVALIDATE to validate the coefficients and model created by TD_ARIMAESTIMATE.
- Use TD_EXTRACT_RESULTS to get the residuals from TD_ARIMAVALIDATE.
- Use TD_SIGNIF_RESIDMEAN to analyze the validate residuals and to determine whether they can be classified as being white noise.
- Retrieve the NULL_HYPOTH results from the TD_SIGNIF_RESIDMEAN output to determine the result of the analysis. A value of 1 means that the series is white noise with a mean of zero. A value of 0 means that the series is not White Noise and has non-zero mean.