|
Activity Number:
|
123
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2003 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Business & Economics Statistics Section
|
| Abstract - #301944 |
|
Title:
|
A Model-Based Criterion for Choosing Filters in the X-12-ARIMA Seasonal Adjustment Program
|
|
Author(s):
|
Yea-Jane Chu*+ and George C. Tiao and William R. Bell
|
|
Companies:
|
Academia Sinica and University of Chicago and U.S. Census Bureau
|
|
Address:
|
Institute of Economics, Taipei, , Taiwan
|
|
Keywords:
|
concurrent adjustment ; mean squared error ; moving average ; seasonal adjustment
|
|
Abstract:
|
The X-12-ARIMA seasonal adjustment program uses empirically developed seasonal and trend moving averages. The user can either specify the particular moving averages or let the program choose them automatically according to some empirical criteria. We propose an alternative approach to filter selection by considering model-based decomposition. More precisely, we propose choosing the filter that minimizes the mean squared error (MSE) when the X-12 adjusted series is used to estimate the underlying nonseasonal component from the model-based decomposition. We show how to compute this MSE for both symmetric and concurrent filters, and for the canonical decomposition and an alternative. The alternative reflects uncertainty about the allocation of white noise between the seasonal and nonseasonal components via a (uniform) prior distribution. Calculations for the "airline model" with various parameter values show in many cases little accuracy is lost in estimating the canonical nonseasonal component by using the best X-12 filter instead of the optimal model-based filter, particularly for concurrent adjustment. The results are less favorable with the uniform prior on white noise allocation.
|
