Dear community,
In the Forecast automation app (Demand planning), we find an unexpected behavior when we set the probability level for the White noise test. Indeed, by lowering the probability (e.g. from 0.9 to 0.6), we expect more white noise to be detected and thus more planning combinations to shift from "Continuous" to "Irregular". Also, having tested with extreme values:
- Probability = 0.000001 --> We'd expect to see all planning combinations considered as having White noise. Instead we get 21 out of 2672
- Probability = 0.999999 --> We'd expect to see no planning combinations with White noise. However the number increased to 149/2672
Our questions:
- Can someone explain the underlying algorithm used for the White noise test, and the exact impact of the probability level?
- Does the presence of patterns (trend and/or seasonality) by definition exclude the presence of White noise? The IBP documentation seems to confirm this, yet we see 7 planning combinations classified as Irregular and showing a trend
Best regards,
Ivan