Partial concurrent adjustment → Estimate regression coefficients + Last outliers

The Partial concurrent adjustment → Estimate regression coefficient + Last outliers strategy means that the ARIMA model, outliers and other regression variables are not re-identified, except the outliers in the last year of the series. Indeed, JDemetra+ has a test procedure to detect outliers.

Additionally to this “outlier-refreshment” procedure, all coefficients of the RegARIMA model are re-estimated. The transformation type remains unchanged.

The example below shows the initial model (left) and the results of the refreshment procedure with the Partial concurrent adjustment → Estimate regression coefficient + Last outliers option (right). Both ARIMA and regression coefficients have been re-estimated. However, the number of estimated coefficients in the revised model is larger than in the initial model because an additional outlier has been identified in the last year of the data span (the level shift in January 2017).

The Partial concurrent adjustment → Estimate regression coefficient + Last outliers revision policy results