|
Activity Number:
|
262
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #307853 |
|
Title:
|
Quantifying and Measuring Revisions in Time-Series Estimates
|
|
Author(s):
|
Duncan Elliot*+ and Craig H. McLaren and Xichaun (Mark) Zhang
|
|
Companies:
|
Office for National Statistics and Office for National Statistics and Australian Bureau of Statistics
|
|
Address:
|
Government Buildings, Cardiff Road, Newport, Gwent, International, NP10 8XG, United Kingdom
|
|
Keywords:
|
revision ; seasonal adjustment ; trend ; decomposition
|
|
Abstract:
|
Time-series estimates, in particular seasonally adjusted and trend estimates, are widely used for analysis. Confidence in these estimates is often related to the size of revisions to past data points. In official statistics, revisions occur for a number of different reasons from changes in methodology to established revision policies for the publishing of data. Information on revisions can vary from simple measurements such as mean absolute percentage error to breakdowns of the source of revisions in the original estimates. We explore methods for providing detailed information on revisions, particularly related to the seasonally adjusted and trend estimates. These approaches are illustrated with real data.
|
