Causal discovery methods extended to handle missing data reliably
Generalizing Identifiability Results in Causal Discovery from Complete Data to Missing Data Settings
Artificial IntelligenceMachine Learning
Summary
Finding cause-and-effect relationships in data is harder when some data is missing. This paper shows conditions where it's still possible to figure out these relationships accurately, even when data isn't complete. The authors provide a way to use methods made for perfect data in situations with missing data, and introduce two algorithms that correctly find causes when certain assumptions apply. Their work helps ensure reliable conclusions even when data has gaps.
What this means in practice
- •For healthcare data teams: Improve causal analysis in patient records with incomplete information using algorithms that account for missing data patterns.
- •For financial data analysts: Analyze cause-effect relations in financial datasets that have missing entries without biasing results by assuming missingness mechanisms.
Authors
Matteo Ceriscioli, Shohei Shimizu, Mallika Gupta, Karthika Mohan
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.