Development of statistical models for data containing heterogeneous subgroups using statistical learning theory and its application to clinical medicine

2023 Fiscal Year Final Research Report

• Project/Area Number: 18K11197
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 60030:Statistical science-related
• Research Institution: Keio University
• Principal Investigator: Hayashi Kenichi 慶應義塾大学, 理工学部(矢上), 准教授 (70617274)
• Project Period (FY): 2018-04-01 – 2024-03-31
• Keywords: 医学統計学 / IDI / ROC / 欠測値 / 因果推論 / 二値回帰モデル / 生存時間解析

Outline of Final Research Achievements

Our research aims to contribute to the development of statistical methods with both predictive power and interpretability for data consisting of heterogeneous subpopulations. The main results are (1) the development of an improved version of the IDI, the odds-IDI, (2) the proposal of statistical analysis methods that take into account various situations in clinical trials, and (3) the proposal of a survival time regression model that takes heterogeneity into account. In (1), the proposed index outperforms existing indices in terms of performance, and the theoretical aspects and interpretability of the proposed index were studied. In (2), statistical analysis methods were studied for the cases where heterogeneity is characterized by missing data and so forth. In (3), a regression model for survival outcomes in the case of a mixture of potentially cured and uncured groups was studied.

Free Research Field

統計科学

Academic Significance and Societal Importance of the Research Achievements

本研究の成果の学術的意義は，従来の統計手法では十分に対応できなかった異質性を含むデータに対する新たな視点からの解析法を提示したことである．予測力と解釈可能性を兼ね備えた新たな統計的手法の開発は，大量・複雑になるデータの特徴を人間が理解する上で重要な意義をもつ．この課題に対し，疫学や臨床試験を想定する諸種のデータについて予測精度のより高い統計モデルを開発し，またそれらのモデルの評価指標を提案した．これらの成果は，様々な分野におけるデータ分析の精度向上と新たな知見の創出に貢献し，社会全体の利益に資する可能性をもつと考えられる．