A Review of the State of the Art in Automated Data Editing and Imputation
Two broad approaches to the automation of data editing and imputation are described. One approach maintains the subject matter specialist as the primary reviewer and corrector of errors but seeks to provide the editor with more powerful tools. A major goal in this approach is to integrate editing with other survey functions such as data collection or data analysis. In the other approach, the computer is being developed as a tool largely to supplant the specialist as a data editor. Software is being developed to analyze edits, choose fields to be corrected, and impute acceptable values. The state of the art in four institutions, Statistics Canada, the U.S. Bureau of the Census, the Netherlands Central Bureau of Statistics, and the National Agricultural Statistics Service is reviewed. Purposes of editing, kinds of editing, its role in addressing nonsampling error, cost, and major issues affecting automation are briefly discussed.
Statistical editing; macro-editing; interactive editing; productivity in survey processing; survey management; survey integration; generalized editing systems.