Estimating the Errors Remaining in a Data Set: Techniques for Quality Control

Publisher Website
Google Scholar
Abstract
This article presents two methods of quantifying the adequacy with which research data have been checked in the process of quality control. In the duplicate performance method, the data operation is carried out twice, independently, and the results are compared; the remaining errors in the data set can be estimated thereby and a confidence limit can be obtained. In the known errors method, the supervisor purposely introduces into a data set known errors similar in form to suspected unknown errors. Then a staff member checks the file; the results yield the number of known errors found and the number of unknown errors found. The method, like the duplicate performance method, allows the accuracy of both workers to be quantified and allows an estimate, with a confidence limit, of the number of as-yet-unfound errors still lurking in the data set.

This publication has 6 references indexed in Scilit: