Score-Based Causal Discovery with Temporal Background Information

2025/02/10 by Tobias Larsen, Claus Thorn Ekstrøm, Larsen, Tobias Ellegaard +3
Computer Science · Decision Sciences · Biochemistry, Genetics and Molecular Biology · #Bayesian Modeling and Causal Inference #Data Quality and Management #Biomedical Text Mining and Ontologies

paper · pdf · doi:10.48550/arxiv.2502.06232

Abstract

Temporal background information can improve causal discovery algorithms by orienting edges and identifying relevant adjustment sets. We develop the Temporal Greedy Equivalence Search (TGES) algorithm and terminology essential for score-based causal discovery with tiered background knowledge. TGES learns a restricted Markov equivalence class of directed acyclic graphs (DAGs) using observational data and tiered background knowledge. To construct TGES we formulate a scoring criterion that accounts for tiered background knowledge. We establish theoretical results for TGES, stating that the algorithm always returns a tiered maximally oriented partially directed acyclic graph (tiered MPDAG) and that this tiered MPDAG contains the true DAG in the large sample limit. We present a simulation study indicating a gain from using tiered background knowledge and an improved precision-recall trade-off compared to the temporal PC algorithm. We provide a real-world example on life-course health data.

Related

• Improving Finite Sample Performance of Causal Discovery by Exploiting Temporal Structure
• Score-matching-based Structure Learning for Temporal Data on Networks
• Less Greedy Equivalence Search
• Do we become wiser with time? On causal equivalence with tiered background knowledge
• Ordering-based Causal Discovery via Generalized Score Matching