Conditional distributional framework for joint modelling of a sensitive attribute and an observed variable under randomized response designs
Journal of the Royal Statistical Society Series C: Applied Statistics · 10.08.2026
DOI: 10.1093/jrsssc/qlag046 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of the Royal Statistical Society Series C: Applied Statistics, online erschienen am 10.08.2026. Verfasst von Shen-Ming Lee und 3 weiteren Autor:innen. Die Arbeit nennt 35 Literaturangaben. Gefördert von: National Science and Technology Council (NSTC) of Taiwan, ROC.
Abstract (Originalsprache)
Randomized response (RR) techniques (RRTs) protect privacy but yield latent sensitive variables. Existing models primarily focus on predicting the sensitive attribute rather than characterizing its association with non-sensitive covariates. We propose a likelihood-based joint modelling framework for a latent binary sensitive variable and an observed non-sensitive variable, conditional on covariates, under an unrelated-question RR design. By parameterizing their joint distribution through a logistic regression model for the sensitive variable and a conditional model for the non-sensitive variable, the framework enables assessment of conditional association between the two variables given covariates. Conditional independence is formally evaluated using likelihood-based tests. Parameter estimation is implemented via an expectation–maximization algorithm that accounts for the RRT mechanism, with established large-sample properties. Simulation studies and an application to the 2012 Taiwan Social Change Survey demonstrate the finite-sample performance and practical utility of the proposed method. All findings are interpreted as conditional associations rather than causal effects.
Förderung
- National Science and Technology Council (NSTC) of Taiwan, ROC
Zitieren
APA
Lee, S., Tran, P., Le, T., & Li, C. (2026). Conditional distributional framework for joint modelling of a sensitive attribute and an observed variable under randomized response designs. Journal of the Royal Statistical Society Series C: Applied Statistics. https://doi.org/10.1093/jrsssc/qlag046
Vancouver
Lee S, Tran P, Le T, Li C. Conditional distributional framework for joint modelling of a sensitive attribute and an observed variable under randomized response designs. Journal of the Royal Statistical Society Series C: Applied Statistics. 2026. doi:10.1093/jrsssc/qlag046
BibTeX
@article{Lee2026,
author = {Lee, Shen-Ming and Tran, Phuoc-Loc and Le, Truong-Nhat and Li, Chin-Shang},
title = {Conditional distributional framework for joint modelling of a sensitive attribute and an observed variable under randomized response designs},
journal = {Journal of the Royal Statistical Society Series C: Applied Statistics},
year = {2026},
doi = {10.1093/jrsssc/qlag046}
}
EUVIMED