Proposal for a robust correction method of publication bias in meta-analysis

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K13270
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 07030:Economic statistics-related
• Research Institution: Yokohama National University
• Principal Investigator: Furukawa Chishio 横浜国立大学, 大学院国際社会科学研究院, 准教授 (30897541)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: メタ分析 / 統計的意思決定論

Outline of Final Research Achievements

This research project has aimed to combine statistical methods of meta-analyses and theoretical models of statistical decisions to explore how to translate statistical analyses into policy making. The project began with the extension of my Ph.D. dissertation . In addition, there were two main sets of progress:
(1) on meta-analysis methods: the effort to synthesize various estimates from studies with different backgrounds rests on the assumption that there is a well-defined common parameter. However, meta-analysis methods have been criticized when the studies substantially differ since the common parameter may not be properly estimated. We develop a new approach to examining the validity of these concerns; (2) on statistical decision models: randomization has been a central principle in statistical inference. However, the expected utility models cannot justify randomization. By augmenting benchmark of expected utility with contemplation costs, we aim to explain the role of randomization.

Free Research Field

経済学、応用統計学、ミクロ理論

Academic Significance and Societal Importance of the Research Achievements

「異なる文脈における推定値をどのように統合するか」、「なぜランダム化が統計推論に役立つか」という問題は伝統的な批判であり、これらの基本的問題への理解を深めることが学術的意義である。また、古典的問題の新しい定式化によって、新しい推定アルゴリズムや十分統計量の公式を導出し、実践的に役立てられることが社会的意義である。