Formalization and automation for statistical causal inference

2018 Fiscal Year Final Research Report

• Project/Area Number: 16K12398
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Statistical science
• Research Institution: Yokohama City University (2018); Chiba University (2016-2017)
• Principal Investigator: Wang Jinfang 横浜市立大学, データサイエンス学部, 教授 (10270414)
• Project Period (FY): 2016-04-01 – 2019-03-31
• Keywords: causal inference / conditional independence / formalization / cain / SSReflect

Outline of Final Research Achievements

(1)Proposed non-inferior tests for image-diagnosis based on cluster data and successfully showed the usefulness of the methods by simulation studies and applying the tests to patients with acute subarachnoid hemorrhage. (2) Proposed random effect models for combining data obtained by multiple raters. New confidence intervals are proposed for both sensitivity and specificity. (3) We carried out formalization of the algebraic system of cain proposed by Wang (2010) for manipulating statistical conditional independence using the proof assistant SSReflect.

Free Research Field

statistics, data science, computer science

Academic Significance and Societal Importance of the Research Achievements

理論と応用の両面において、統計的因果推論は極めて重要な問題である。本研究では、この問題を正面から挑戦し、独自の代数系を提案し、それに基づく条件付き独立性の形式化を行った。これにより、条件付き独立性に関する操作の半自動化・自動化の可能性を開いた。