Online Program

Metrics for Selective Editing
*Joseph Keating, Central Statistics Office 

Keywords: Selective Editing,Non-Response

In selective editing, edits are given a score according to some metric and only those edits with an edit score above a threshold are investigated. The aim is to reduce the costs of traditional editing approaches while maintaining the quality of statistical results. The skewed nature of enterprise surveys makes them especially suitable for such an approach. Edit scores usually combine some measure of the Suspicion and the Impact associated with a value. This paper discusses the development and use of a selective editing approach in editing a historic dataset where editing resources were limited. The paper shows how the measure of Impact can be calculated directly, and discusses various options for combining it with the measure of Suspicion. It discusses the advantages and disadvantages of this approach when compared with other approaches to selective editing, and outlines how it can be extended to more general situations. Finally it looks at how an adaptation of this approach can be used to identify the most important current non-responding enterprises and proposes a measure for the benefit to be gained from obtaining a response from a particular enterprise.