15/04/2009 Note: PDF Replaced. A correction was made to equation (2) on page 4. No other changes have been made.
Selective editing is not new. The Australian Bureau of Statistics (ABS) calls its version significance editing. It has been used in various guises for the last 10 years in an ad-hoc manner (except for some small groups of economic and agricultural collections). This paper looks to extend the application of significance editing to more difficult situations. This paper seeks feedback on the approaches taken and instructions on worthwhile future directions.
The paper explains how significance editing has been applied to situations where:
(i) there are a large number of key variables;
(ii) no data for before editing and after editing has been saved;
(iii) there are one-off surveys and surveys without historical data;
(iv) surveys which are unable to reasonably predict future estimates;
(v) surveys where responses are not able to be predicted suitably; and
(vi) surveys have editing strategies that need more than identification of providers.
The paper also looks into best use of significance editing and best alignment of the various editing streams.