Measurement of Type I and Type II Record Linkage Error

• Activity Number: 469 - Using and Linking Administrative and Auxiliary Data
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Government Statistics Section
• Author(s): Dean Resnick* and Jana Lynn Asher
• Companies: National Opinion Research Center (NORC) and Slippery Rock University
• Keywords: record linkage; administrative records ; type I error; type II error; empirical Bayes

Abstract

Measurement of Type I and II error associated with record linkage is an ongoing research topic. Most authors have either relied on a "gold standard" identifier for measuring error, post-linkage data validation through finding implausible or impossible linked data (e.g., two or more death records), sensitivity analysis, or comparison of linked and unlinked data to explore biases. This paper presents an alternative method for measurement of both Type I and Type II error through a direct estimation of the probability of a match, dependent on m and u probability values determined during the matching process. Simulated data are used to show the accuracy of the method in estimating both types of error.