Setting Thresholds for Selective Edit: A Methodological Approach Using Process History from Establishment Surveys

• Activity Number: 66
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Joseph Kosler*+
• Companies: USDA NASS RDD
• Keywords: selective edit ; significance edit ; automation ; development ; judgment sample ; SignEdit

Abstract

The National Agricultural Statistics Service (NASS) is introducing an automated system for significance editing (SignEdit System) to the production environment for establishment surveys. The system is integrated with Blaise 4.73 to provide for interactive edit of judgment samples based on the sort order of a unit score assigned to each farm report. To expedite selection of the judgment sample, methodology for a selective edit threshold is being considered. The threshold may be vector-valued to support subdomain differences. The methodology was adapted from a trusted procedure in process control for biotech vaccines into a new procedure for gauging subdomain-level threshold values. The equivalent of two years process history was used. Operational weight-adjustment procedures were considered. We were able to capture intermediate adjustment factors from summary calculations and characterize their relative impact to the unit score. Methodology for setting and maintaining threshold values is under development for a smaller survey with plans to scale up to production for a larger survey. Operationalization of any threshold is contingent upon their reliability and performance.