Coherent and robust Bayesian inference under informative sampling via sample-level weight models
Abstract
We develop two complementary Bayesian procedures for analytic inference under informative sampling. The first constructs a coherent likelihood by modeling the sample-level distribution of the design weight given the study variable and recovering the conditional inclusion probability through the identity of Sverchkov and Pfeffermann (2004); it is the Bayesian counterpart of a recently proposed conditional-likelihood estimator, attains the model-based efficiency bound under the correctly specified sample-level joint model, and can be biased otherwise. The second uses the same weight model only as input to a Neyman-orthogonalized analog of the design-weighted score and applies the loss-likelihood bootstrap; the resulting credible intervals are asymptotically sandwich-correct under Poisson sampling, and the posterior remains centered at the truth under misspecification of the weight model, although the working model can still affect the first-order variance. A beta regression on the population-level inclusion probability, inducing a beta prime sample-level weight distribution, serves as the parametric default. A data-adaptive implementation of the second procedure estimates the efficient factor π̅(x,y)=1/Ep(W| x,y) by a design-weighted Gamma/log-link regression and uses it in a joint one-step loss-likelihood bootstrap for the analytic parameter; its validity requires no rate conditions on the nuisance fit. We illustrate both procedures with simulations and Canadian Workforce data.
Related
- Saturation in G: simple & robust causal inference in cluster randomized trials with informative cluster sizes
- Likelihood-Free Schrödinger--Föllmer Sampling for Bayesian Inference with Fixed-Proposal Reuse
- A computationally-tractable measure of global sensitivity for sampling-based Bayesian inference
- Calibration-Free Robust Bayesian Inference for Discrete Unnormalized Models
- Robust Bayesian Inference for Unnormalized Models with Mixed-Domain Data
BibTeX
@misc{morikawa2026coherent,
title = {{Coherent and robust Bayesian inference under informative sampling via sample-level weight models}},
author = {Kosuke Morikawa and Jae Kwang Kim and Won Chang},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.16225},
eprint = {2609.16225},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.16225},
}
RIS
TY - UNPB
TI - Coherent and robust Bayesian inference under informative sampling via sample-level weight models
AU - Kosuke Morikawa
AU - Jae Kwang Kim
AU - Won Chang
PY - 2026
DA - 2026/09/14
PB - arXiv
DO - 10.48550/arxiv.2609.16225
UR - https://arxiv.org/abs/2609.16225
AB - We develop two complementary Bayesian procedures for analytic inference under informative sampling. The first constructs a coherent likelihood by modeling the sample-level distribution of the design weight given the study variable and recovering the conditional inclusion probability through the identity of Sverchkov and Pfeffermann (2004); it is the Bayesian counterpart of a recently proposed conditional-likelihood estimator, attains the model-based efficiency bound under the correctly specified sample-level joint model, and can be biased otherwise. The second uses the same weight model only as input to a Neyman-orthogonalized analog of the design-weighted score and applies the loss-likelihood bootstrap; the resulting credible intervals are asymptotically sandwich-correct under Poisson sampling, and the posterior remains centered at the truth under misspecification of the weight model, although the working model can still affect the first-order variance. A beta regression on the population-level inclusion probability, inducing a beta prime sample-level weight distribution, serves as the parametric default. A data-adaptive implementation of the second procedure estimates the efficient factor π̅(x,y)=1/E_p(W| x,y) by a design-weighted Gamma/log-link regression and uses it in a joint one-step loss-likelihood bootstrap for the analytic parameter; its validity requires no rate conditions on the nuisance fit. We illustrate both procedures with simulations and Canadian Workforce data.
ER -