Any change in the data structure introduces a break in the time series. This is called a discontinuity.
Points of discontinuity in a time series must be identified and explained.
Substantial revisions to time series should, if possible, provide:
- consistent back-data series
- analysis of the differences between the old and revised series
- explanation of the effect on any previously published commentary or interpretation.
Users must be advised of substantial conceptual and methodological changes before statistics are released based on the new methods.
Backcasting
A number of statistical modelling methods are available to produce estimates for past time periods under the new measurement conditions. This is called backcasting.
The method chosen depends on the needs of the users, and the availability of information and resources.
Guidance
- Principles and Protocols for producers of Tier 1 statistics (New Zealand’s most important official statistics):
- Protocol 1 – Quality
- Element 7 – Consistency
- Protocol 1 – Quality