S. Díaz Aranda, J. M. Ramírez, J. Aguilar, A. Fernández Anta, R. E. Lillo

The Network Scale-up Method (NSUM) is a relatively recent survey-based approach for estimating hard-to-reach populations based on the aggregate data of the respondents’ acquaintances. However, NSUM is prone to biases arising from participant behavior, particularly due to sensitivity to small deviation because of the reliance on sample means. This work investigates the use of eight robust proposals for the two classical NSUM to mitigate errors caused by misreporting, contamination, barrier effects, prevalence, skewness, and tail length. We evaluate our proposals through simulation experiments using synthetic networks that mimic real-world social structures, as well as real data. The results demonstrate that classical NSUM estimators perform poorly in contaminated settings. While robust methods generally improve performance, some degrade under barrier effects. Additionally, distortions caused by low prevalence significantly influence the selection of the most effective robust estimator

Keywords: network scale-up method; aggregated relational data; robust estimators; adaptive procedures; M-estimators

Scheduled

AMC2 Robust Methods
June 10, 2025  5:10 PM
Auditorio 1. Ricard Vinyes


Other papers in the same session

Robust model-based clustering via cellwise approach

L. A. García-Escudero, G. Greselin, A. Mayo-Iscar, G. Zaccaria

Shape-constrained functional regression with penalized splines

M. Durbán, V. Guerrero, C. Molero-Río, J. Secilla Martínez


Cookie policy

We use cookies in order to be able to identify and authenticate you on the website. They are necessary for the correct functioning of it, and therefore they can not be disabled. If you continue browsing the website, you are agreeing with their acceptance, as well as our Privacy Policy.

Additionally, we use Google Analytics in order to analyze the website traffic. They also use cookies and you can accept or refuse them with the buttons below.

You can read more details about our Cookie Policy and our Privacy Policy.