Development of Causal Inference Methods Based on Sparse Modeling and Bayesian Decision Theory

2023 Fiscal Year Final Research Report

• Project/Area Number: 19K12128
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: Waseda University
• Principal Investigator: Horii Shunsuke 早稲田大学, データ科学センター, 准教授 (00552150)
• Project Period (FY): 2019-04-01 – 2024-03-31
• Keywords: 統計的因果推論 / 構造的因果モデル / 潜在反応モデル / 統計的決定理論 / ベイズ統計学 / スパースモデリング

Outline of Final Research Achievements

This study aimed to optimize statistical causal inference within the framework of statistical decision theory to avoid erroneous insights in data analysis. First, a Bayesian decision theory-based method for estimating causal effects was proposed and its effectiveness was demonstrated through simulations and experiments with real data. Additionally, efficient approximation algorithms using MCMC and variational Bayesian methods were developed to address computational complexity issues, enabling high-precision estimation even with small-scale data. Furthermore, methods for causal inference using instrumental variables and for estimating conditional average treatment effects were proposed, demonstrating superiority over conventional methods. The research outcomes were presented at top conferences such as AISTATS and AAAI, contributing to the fields of statistics and artificial intelligence.

Free Research Field

統計的因果推論

Academic Significance and Societal Importance of the Research Achievements

本研究の学術的意義は、統計的因果推論を統計的決定理論の枠組みで最適化し、従来の方法よりも精度の高い因果効果の推定を可能にした点にある。これにより、因果推論の理論体系が深化し、多様なデータ分析における新たな知見の創出が期待される。社会的意義としては、データに基づく意思決定の精度向上が挙げられる。特に医療や経済学などの分野で、因果関係の正確な把握に基づく政策や治療法の最適化が可能となり、公共の福祉や経済の発展に寄与することが期待される。