Back to search

University of Southampton

Automated probabilistic record linkage without classification for dual system population size estimation

Abstract

dc:description.abstract

Population size estimation from two incomplete surveys, known as dual system estimation, requires to know which of the population elements are simultaneously captured in both of the surveys, the task imperfectly accomplished by the means of record linkage. In this thesis we explore the conceptual closeness of the fields of probabilistic record linkage and dual system estimation, and develop methods for the population size estimation, called linkage free dual system estimation, that seamlessly integrate probabilistic record linkage and dual system estimation. Unlike many existing record linkage approaches, the one developed in this thesis is purely estimation-based and does not classify records into links and non-links. It also does not require clerical resolution of possible links. In order to theoretically justify the linkage free dual system estimation method, we revisited certain problematic aspects of probabilistic record linkage and proposed a different approach conceptualizing record linkage models. This conceptualization takes into account a very specific sampling mechanism behind record linkage tasks. It also allows analysis of certain properties and limitations of parameter estimation in linkage models. We also introduce a special case of data blocking that bridges the gap between record linkage data and estimation with these data. Special attention is paid to between-variables associations in the outcomes obtained by comparing the values of linkage variables. We also assess linkage models for identifiability using a variety of methods from the field of algebraic statistics. We demonstrate that in situations where the data in both surveys are collected for the same geographical clusters, the linkage free dual system estimation is feasible and can yield outputs of similar quality to the regular classification approaches that involve clerical interventions. We also develop accompanying variance estimation methods, and these methods rely on less restrictive assumptions than existing methods. All developments are undertaken within the frequentist paradigm.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D.
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Southampton
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Racinskij, Viktor
Advisors dc:contributor.advisor
  • Smith, Paul
  • Van Der Heijden, Peter
ORCID person:identifier.orcid

Rights

Language dc:language
English, eng

Chain of custody

source
Harvested from
University of Southampton
Base URL
eprints.soton.ac.uk/cgi/oai2
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Racinskij, Viktor. Automated probabilistic record linkage without classification for dual system population size estimation. doctoral thesis, University of Southampton, 2024.