2006 Meeting of the SSC, London, Ontario

Activities Sponsored by the Survey Methods Section

Workshop on Survey Methods, May 28, 2006

Introduction to Record linkage

* Patricia Whitridge, Elections Canada
* Karla Fox, Statistics Canada

Record linkage is simply the integration of information from two independent sources. It has become increasingly useful in maintaining electoral information, health care administration, demographic studies, and in medical research. It has enabled administrators and researchers to maintain a database, link to files (i.e. administrative), link to survey data, and/or link to clinical data. Under this framework, records are linked on the basis of common data. Records from the two or more sources that are believed to relate to the same individual are matched in such a way that they may then be treated as a single record for that individual. A match in a deterministic linkage is made when a sufficient number of identifiers agree exactly between two records. In probabilistic record linkage, the comparison or matching algorithm yields for each record pair, a probability or “weight” which indicates the likelihood that record pairs relate to the same entity.

This course will cover the general principles of record linkage, including data preparation, deterministic linkage, probabilistic linkage, and linkage evaluation. Different principles will be illustrated with data from real life examples and there will be a general discussion on record linkage software packages currently available and how to evaluate them for your particular needs.

Invited Sessions

Special Invited Session: Record linkage and counterterrorism

Organizer: Patricia Whitridge, Elections Canada

* Michael Larsen, Iowa State University

Invited Session I: Analysis of longitudinal complex survey data using marginal modelling and GEE: theory and practice

Organizer: Susana Rubin-Bleuer, Statistics Canada

* Georgia Roberts, Statistics Canada
* Abdelnasser Saidi, Statistics Canada
* Punam Pahwa, Dept. of Community Health and Epidemiology
* Sunita Ghosh, University of Saskatchewan

Invited Session II: Analyzing data from international surveys

Organizer: Diane Stukel, UNESCO

* Scott Murray, UNESCO
* Miles Corak, Statistics Canada
* Edilberto Loaiza, UNICEF

Invited Session III: Latent variable and multilevel modelling using survey data

Organizer: Changbao Wu

* Sophia Rabe-Hesketh, University of California, Berkeley
* Eric Lacourse, University de Montréal
* Milorad Kovacevic, Statistics Canada

Invited Session IV (Cosponsored with the NPCDS): Some current developments in the analysis of complex survey data.

Organizer: M. Kovacevic and C. Wu

* Ivan Carrillo-Garcia, University of Waterloo
* Xiaojian Xu, University of Alberta
* Chunfang Lin, Simon Fraser University

Case Study: Record linkage

Organizer: Patricia Whitridge, Elections Canada, and Karla Fox, Statistics Canada

This study introduces the student to different record linkage methods. Data linkage falls into two large categories: statistical matching and exact matching. In many research and administrative situations the need arises to link two or more data sources together. The student assignment is given in progressively complex steps. First the student will use record linkage techniques to ‘update’ a population registry. Next the student can estimate linkage and error rates.

2007 Joint Statistical Meetings, Minneapolis

Invited Session Sponsored by SSC and Organized by the Survey Methods Section

Balanced Sampling

Organizer: P. Lavallée

* Yves Tillé, Université of Neuchâtel
* Jean-Claude Deville, CREST/ENSAI
* Marc Christine, INSEE
* Wisner Jocelyn, Statistics Canada