返回Google圖書搜尋

New Developments in the X-11-ARIMA

出版	Statistics Canada, Methodology Branch, Time Series Research and Analysis Division, 1987
URL	http://books.google.com.hk/books?id=DfDGtAEACAAJ&hl=&source=gbs_api

註釋The X-11-ARIMA Seasonal Adjustment computer program (Dagum 1980) performs three basic functions, namely, (1) forecasting; (2) seasonal adjustment and (3) composition of original and seasonally adjusted series. Major developments are being carried out in two key functions in order to increase the efficiency of the method and its ability to handle adequately a larger number of series. New developments related to the forecast functions are (i) a new set of built-in ARIMA models; (ii) variable forecasting horizons; (iii) backcasting only series shorter than seven years; (iv) new acceptance criteria of fitting and extrapolation for the built-in ARIMA models; (v) determining the optimal forecast horizon as a function of both the amount of noise in the series and its trend-cycle pattern; (iv) introduction of other extrapolation methods. Developments related with the seasonal adjustment function include (i) estimation of Easter effect; (ii) estimation of stochastic trading-day variations; (iii) new replacement of extreme values; (iv) increasing the accuracy of the end-weights of the five Henderson trend-cycle filters; and (v) new diagnostic tools. The main purpose of this paper is to provide a summary of each one of these major developments and their impact on the two corresponding basic functions of X-11-ARIMA.