Survey Methods Section Workshop

Record Linkage Methods: Theory and Application


Abdelnasser Saïdi, Statistics Canada

June 14, 9:00 am to 4:00 pm


Record linkage is being used more and more frequently in many fields of study: the electronic maintenance of information registries, health and epidemiology, and the elimination of duplicates in order to create survey frames.


Record linkage consists of matching records that contain information on persons, businesses, or households when no unique identifier is available. Within this framework, the records are linked according to which data fields they have in common. This workshop focuses in detail on probabilistic record linkage based on the Fellegi-Sunter theoretical model, in which pairs of records are classified as linked, unlinked, or possibly linked which then have to be resolved manually. We present details of the difficulties with the preparation of data for linkage. At Statistics Canada, probabilistic record linkage is used frequently. To facilitate its use, Statistics Canada has implemented its own probabilistic record linkage program, called “G-Link” that is based on the Fellegi-Sunter algorithm. This workshop will provide a complete review of the methodology used in conjunction with concrete examples developed using G-Link. Finally, there will be a discussion on recent research and innovations in this area.


Presentation


Abdelnasser Saïdi is a senior methodologist and head of the Record Linkage Resource Center at Statistics Canada. He worked in many areas including sampling, analysis of survey data and record linkage. He is a Phd in Statistics from University of Grenoble (France).