Population Size Estimation for Respondent-Driven Sampling and Capture-Recapture: A Unifying Framework

TL;DR

This paper presents a unifying framework for estimating hidden population sizes by modeling respondent-driven sampling as a capture-recapture process, offering new methodologies and performance insights.

Contribution

It introduces a novel perspective by treating RDS data as a capture-recapture experiment and proposes a new estimation methodology for population size.

Findings

The proposed method effectively estimates population size from RDS data.

Performance remains robust under deviations from classical capture-recapture assumptions.

The framework unifies RDS and capture-recapture approaches for hidden populations.

Abstract

This paper deals with the estimation of population sizes for respondent-driven sampling (RDS), a variant of link-tracing sampling that leverages social networks over a number of waves to recruit individuals from hidden populations. The RDS process is mostly controlled by individual participants who might report on recruitment proposals, or nominations, that they have received or given. By considering all nominations given or received over a time period, one can create a capture-recapture dataset in which units are individuals who have received at least one nomination and capture occasions are either time intervals or recruitment waves, with the goal of estimating the size $N$ of the hidden population. In this paper, we argue that the underlying process that generated the RDS nomination data is that of a capture-recapture experiment. We then proposed a methodology for the estimation of the population size and investigated its performance against departures from classical capture-recapture assumptions.