Counting the Uncounted: Methodological Extensions in Multiple Systems Estimation

Research output: ThesisDoctoral thesis 1 (Research UU / Graduation UU)

Abstract

Multiple systems estimation (MSE) is an estimation method to estimate the number of unobserved population units, in case a population is observed by two or more (incomplete) samples of this population. To provide asymptotically unbiased estimates, MSE depends on a number of assumptions and conditions that are often violated in practice. When one of the assumptions or conditions is violated, there generally is a method available in literature that allows the statistician to correct for this violation. However, when more than one assumption or condition is violated, it is not always clear how these correction methods can be combined. To be more specific, an MSE assumption that is often violated, is the assumption of sample independence, which will lead to a biased estimate when only two samples are used. When more than two samples are used, this assumption can be relaxed, so the use of three or more samples is advisable. However, some of the available correction methods for other violations of assumptions or conditions, such as the assumption of perfectly identifiable population units and the condition of large enough samples, are generally developed for the case of two (independent) samples. Furthermore, when multiple samples are required, it may be that some of these samples are available with delay, while an early estimate is desired. Therefore, in this dissertation we propose methods that generalise the finite sample correction method by Chapman (1951) in case of three or more samples (Ch. 2), improve and extend the so-called linkage error correction method by Ding & Fienberg (1994, Ch. 3 & 4), and proposes two different approaches (Ch. 5 & 6) that allow the statistician to obtain early estimates, where Ch. 6 proposes a new method that is based on samples that become available periodically over time.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Utrecht University
Supervisors/Advisors
  • van der Heijden, Peter, Supervisor
  • Bakker, Bart, Supervisor, External person
Award date22 Nov 2024
Publisher
DOIs
Publication statusPublished - 22 Nov 2024

Keywords

  • population size estimation
  • multiple systems estimation
  • capture-recapture
  • Mark and recapture
  • Log-linear model
  • Finite sample bias
  • Hard-to-reach groups
  • official statistics
  • time series analyses
  • linkage errors

Fingerprint

Dive into the research topics of 'Counting the Uncounted: Methodological Extensions in Multiple Systems Estimation'. Together they form a unique fingerprint.

Cite this