Fast approaches for Bayesian estimation of size of hard-to-reach populations using Network Scale-up

TL;DR

This paper introduces fast Bayesian estimation techniques, including Gibbs sampling and Monte Carlo methods, to efficiently estimate the size of hard-to-reach populations using the Network Scale-up method, demonstrated on drug user data from Brazil.

Contribution

It presents novel computational approaches for Bayesian estimation in the Network Scale-up method, improving speed and efficiency over traditional methods.

Findings

Gibbs sampling effectively estimates population size.

Monte Carlo approach provides accurate results.

Methods applied successfully to real-world data.

Abstract

The Network scale-up method is commonly used to overcome difficulties in estimating the size of hard-to-reach populations. The method uses indirect information based on social network of each participant taken from the general population, but in some applications a fast computational approach would be highly recommended. We propose a Gibbs sampling method and a Monte Carlo approach to sample from the random degree model. We applied the abovementioned analytical strategies to previous data on heavy drug users from Curitiba, Brazil.