Generalizing Identifiability Results in Causal Discovery from Complete Data to Missing Data Settings
Abstract
While causal discovery methods have matured for complete data, such algorithms for handling missing values remain limited. We establish general sufficient conditions under which causal graphs remain identifiable (i.e., uniquely determined from the available data distribution) under various missingness categories, including Missing Not At Random, and provide a framework to extend correctness guarantees from causal discovery algorithms for complete data to missing data settings. Our results show that identifiability of causal graphs with no unobserved confounders is preserved when (1) missingness mechanisms do not causally interact, (2) no variable directly causes its own missingness, and (3) either faithfulness holds for missingness mechanisms, or they follow generalized linear models. Guided by these theoretical results, we introduce two asymptotically correct algorithms for Additive Noise Models (ANMs) that recover the underlying causal graph and provide consistent parameter estimates.
Related
- Causal Discovery beyond Conditional Independence: Identifiability, Algorithms, and Statistical Inference for Directed Acyclic and Cyclic Graphs
- Effective Results for Foliations on Smooth Projective Complete Intersection Surfaces
- Admissibility and Complete Classes for False Discovery Rate Control with E-values
- Maximum Intersecting Families of Submultisets: A Complete Solution over All Ranks and Exact Results at Fixed Rank
- Generalizing the AA Problem for Finite Ordered Sets
BibTeX
@misc{ceriscioli2026generalizing,
title = {{Generalizing Identifiability Results in Causal Discovery from Complete Data to Missing Data Settings}},
author = {Matteo Ceriscioli and Shohei Shimizu and Mallika Gupta and Karthika Mohan},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.19211},
eprint = {2610.19211},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.19211},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Generalizing Identifiability Results in Causal Discovery from Complete Data to Missing Data Settings
AU - Matteo Ceriscioli
AU - Shohei Shimizu
AU - Mallika Gupta
AU - Karthika Mohan
PY - 2026
DA - 2026/10/13
PB - arXiv
DO - 10.48550/arxiv.2610.19211
UR - https://arxiv.org/abs/2610.19211
AB - While causal discovery methods have matured for complete data, such algorithms for handling missing values remain limited. We establish general sufficient conditions under which causal graphs remain identifiable (i.e., uniquely determined from the available data distribution) under various missingness categories, including Missing Not At Random, and provide a framework to extend correctness guarantees from causal discovery algorithms for complete data to missing data settings. Our results show that identifiability of causal graphs with no unobserved confounders is preserved when (1) missingness mechanisms do not causally interact, (2) no variable directly causes its own missingness, and (3) either faithfulness holds for missingness mechanisms, or they follow generalized linear models. Guided by these theoretical results, we introduce two asymptotically correct algorithms for Additive Noise Models (ANMs) that recover the underlying causal graph and provide consistent parameter estimates.
ER -