|
The progresses in information technology have fostered the availability of daily and weekly time series. The seasonal adjustment of high-frequency time series poses several challenges. First of all the seasonal period of the annual cycle is neither constant nor an integral. Secondly, in order to accommodate complex seasonal patterns many individual effects might be required. Thirdly, the need for robust methods is reinforced by the fact that the effects of outlying observations is not smoothed by temporal aggregation and that they are relatively more frequent.
Moving average methods (X-13ARIMA-SEATS) and ARIMA-based methods (TRAMO-SEATS) are recommended by Eurostat and are implemented in JDemetra+, the European software for seasonal adjustment. These methods only deal with monthly and quarterly series; they estimate first the outliers and calendar effects using a Reg-ARIMA model and then decompose the residual of the model into trend-cycle, seasonality and irregular component.
We will present a first implementation of a Tramo-Seats algorithm and a X-12 algorithm to seasonally adjust high frequency data with multiple and non-integer seasonalities.
|
