New statistical model enables causal DAG identification from count data

Researchers have introduced a new statistical framework called the Poisson thinning structural equation model (PT-SEM) for identifying causal directed acyclic graphs (DAGs) from observational count data. This model extends the existing Poisson branching structural causal model (PB-SCM) by utilizing Poisson thinning and allowing for various count distributions for exogenous variables. The PT-SEM framework establishes the identifiability of the causal DAG, thinning coefficients, and exogenous distributions under specific conditions, offering full identifiability in certain scenarios. A structure learning algorithm based on dynamic programming and BIC scores has also been developed, demonstrating strong performance in simulations and a real-world application. AI

IMPACT Provides a new methodological tool for causal inference in AI and machine learning research involving count data.

RANK_REASON Academic paper introducing a new statistical methodology. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Poisson branching structural causal model
• Poisson thinning structural equation model
• SIG (Company)

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →