Score-Based Methods for Selecting Direct Causes in Reweighted Distributions

Abstract

Scoring methods based on the Bayesian information criterion (BIC) are commonly used for model selection tasks such as choosing features in a regression model or learning directed acyclic graphs from data. In certain model classes, model selection based on the BIC is consistent. However, since the BIC is based on the observed likelihood function, it does not apply directly to reweighted distributions arising in causal inference, missing data, and domain shift applications. Here, we propose a generalized version of the BIC that allows for model selection in reweighted distributions. We prove its corresponding consistency property and demonstrate how it can be used for selecting direct causes of an outcome variable in a marginal structural model. Our proposed method accounts for scenarios with unmeasured confounders and where the weights must be estimated from data. Through simulation studies, we demonstrate the asymptotic properties of the reweighted BIC and compare it with alternative methods for model selection.

Direct Simulation of Poisson and Newton Kernel Distributions
Lukas Sablica, Kurt Hornik
arXiv23 Oct 2026PreprintarXiv licenceFull textStatisticsarXiv
Can the All of Us sample be reweighted to mirror a nationally representative sample? A comparison of mortality predictors
Jingxuan Wang, Peter Buto, Erin L. Ferguson +9
Epidemiology (Cambridge, Mass.)25 Jun 2026Peer-reviewedLicence not statedFull textStatisticsPMC
A Kinetic Theory Approach To Ordered Fluids: Direct Simulation Monte Carlo Methods
José A. Carrillo, Patrick E. Farrell, Andrea Medaglia +1
arXiv25 Sep 2026PreprintCC BY 4.0Full textMathematical physicsarXiv
Costate Estimation and Covector Mapping for Bernstein Polynomial Direct Methods in Optimal Control
Venanzio Cichella, Isaac Kaminer, Claire Walton +2
arXiv1 Nov 2026PreprintarXiv licenceFull textOptimisation & controlarXiv
Neural Sampling with Reweighted Normalizing Flows via the Wasserstein--Fisher--Rao JKO Scheme
Chenguang Duan, Johannes Hertrich, Gabriele Steidl
arXiv7 Oct 2026PreprintarXiv licenceFull textNumerical analysis & scientific computing ProbabilityarXiv

Cite

BibTeX RIS

BibTeX

@misc{chen2026scorebased,
  title = {{Score-Based Methods for Selecting Direct Causes in Reweighted Distributions}},
  author = {Jacob M. Chen and Ilya Shpitser},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.09100},
  eprint = {2610.09100},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.09100},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Score-Based Methods for Selecting Direct Causes in Reweighted Distributions
AU  - Jacob M. Chen
AU  - Ilya Shpitser
PY  - 2026
DA  - 2026/10/06
PB  - arXiv
DO  - 10.48550/arxiv.2610.09100
UR  - https://arxiv.org/abs/2610.09100
AB  - Scoring methods based on the Bayesian information criterion (BIC) are commonly used for model selection tasks such as choosing features in a regression model or learning directed acyclic graphs from data. In certain model classes, model selection based on the BIC is consistent. However, since the BIC is based on the observed likelihood function, it does not apply directly to reweighted distributions arising in causal inference, missing data, and domain shift applications. Here, we propose a generalized version of the BIC that allows for model selection in reweighted distributions. We prove its corresponding consistency property and demonstrate how it can be used for selecting direct causes of an outcome variable in a marginal structural model. Our proposed method accounts for scenarios with unmeasured confounders and where the weights must be estimated from data. Through simulation studies, we demonstrate the asymptotic properties of the reweighted BIC and compare it with alternative methods for model selection.
ER  -

Abstract

Related

Cite