Recommendations

The Recommendation functionality by the CompRes plugin helps with the decision of whether to keep or not to keep using the previously forecast seasonal component. This feature is useful when working with the controlled current seasonal adjustment regime. It compares two seasonal adjustment results. First, the results obtained when using a seasonal component that has been forecast in the past. Second, the results obtained when using the seasonal component that has been estimated right now by JD+ given the latest available data.

Outcome

Three different output results are possible:

Recommendation Meaning
Check Check the specification of the SA-Item.
Update Switch to the new components.
Keep Continue with the current seasonal and calendar components.


The recommendations are color-coded similar to a traffic light: red for Check, yellow for Update and green for Keep.
If no such recommendation can be given the output result Unknown is displayed in red.


Specifications

For each SA-Item specifications can be changed in the properties sheet ($\text{Window} \rhd \text{Properties}$). To display the recommendation specification select the SA-Item(s), right mouse click and select $\text{CompRes Recommendations} \rhd \text{Add MetaData}$. To reset an SA-Item’s specification to the default setting select $\text{CompRes Recommendations} \rhd \text{Reset MetaData}$. If not specified, a default value will be applied.

Name Description Default
compres.nd8 Number of last periods for which the SI-ratios are used for the comparison. 3
compres.tolgrowth Acceptable absolute deviation between new and old growth rates. 1.0
compres.ngrowth Number of last periods whose growth rates are used for the large movement detection. 10
compres.trim Defines the quantiles for the growth rates to be used in the large movement detection. 0.05


Special Cases

If you select one or both the checkboxes under $\text{Tools} \rhd \text{Options} \rhd \text{Demetra} \rhd \text{CompRes} \rhd \text{Recommendations: Special Cases}$ it might affect the recommendation.
In case of constant seasonal factors (current), the recommendation can be overwritten so that it will always be Keep.
In case the calendar factor (current) is missing, seasonal adjustment without a calendar component is assumed.
Otherwise the recommendation Unknown will be given.