Limitations and possibilities of statistical causal inference using proxy variables in higher education research.

2023 Fiscal Year Final Research Report

• Project/Area Number: 22K20217
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund
• Review Section: :Education and related fields
• Research Institution: Hiroshima City University
• Principal Investigator: Nakao Ran 広島市立大学, 企画室, 特任助教 (80965434)
• Project Period (FY): 2022-08-31 – 2024-03-31
• Keywords: 因果推論 / 代理指標 / 部分的統制

Outline of Final Research Achievements

This research has been conducted with the aim of quantifying bias in statistical causal inference using proxy indicators and developing sensitivity analysis methods. We have studied what kind of proxy indicators can lead to precise statistical causal inference by focusing on how measurement error arises, quantifying the bias when using proxies for the dependent variable, trade-offs with missing proportions, and when daring to control for intermediate variables as proxies. In each case, it was not possible to say that the exact proxy would reduce the bias the most, but it was necessary to conduct sensitivity analysis by creating variables and measuring variables in such a way that the bias would be reduced the most in relation to the other variables.

Free Research Field

計量社会学

Academic Significance and Societal Importance of the Research Achievements

本研究の成果から正確な測定が精緻な推定に必ずしも繋がるとは言えず，統計的因果推論の視点から正確な測定を捉えるという新たな見方を提案できた。この新しい視点に基づく，正確な測定手法，調査手法の開発が今後必要になることを示唆した。