Development of statistical analysis for clarifying cause-effect relationships based on statistical causal models

2018 Fiscal Year Final Research Report

• Project/Area Number: 15K00060
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Statistical science
• Research Institution: Yokohama National University (2017-2018); The Institute of Statistical Mathematics (2015-2016)
• Principal Investigator: KUROKI Manabu 横浜国立大学, 大学院工学研究院, 教授 (60334512)
• Project Period (FY): 2015-04-01 – 2019-03-31
• Keywords: 統計的因果推論

Outline of Final Research Achievements

In this research, I considered the evaluation problem of causal effects to clarify causal mechanism, and I showed that intermediate variables improve the estimation accuracy of causal effects compared the covariate adjustment in some situations. In addition, based on structural causal models, I provided the mathematical formulations of some concepts used in practical science such as the traffic conflict. Based on the formulations, I clarified the relationships between these concepts and "the probabilities of causations", and provided some identification conditions.

Free Research Field

統計科学

Academic Significance and Societal Importance of the Research Achievements

因果メカニズムの解明において，中間変数が重要な役割を果たすことは認識されていたが，本研究により，中間変数の情報を用いることで因果効果の推測精度も改善できる可能性があることが明らかとなった．また，これまでの研究においては，交通コンフリクトなどの実質科学的概念がどのような形式で因果推論と結びつくのかが明らかにされることはなかった．本研究では，これらの概念を因果推論の立場から定式化することで，因果的に解釈できるための要件が明らかとなった．