Absolute model evaluation measures and evaluations of predictive densities

2021 Fiscal Year Final Research Report

• Project/Area Number: 18K11200
• 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: Tokyo University of Science
• Principal Investigator: Kurosawa Takeshi 東京理科大学, 理学部第一部応用数学科, 准教授 (80582303)
• Co-Investigator(Kenkyū-buntansha): 江島 伸興 久留米大学, 医学部, 客員教授 (20203630)
• Project Period (FY): 2018-04-01 – 2022-03-31
• Keywords: モデル評価 / 予測分布

Outline of Final Research Achievements

Generalized linear models that are generalizations of the linear models include various types of models such as logistic models, multinomial logit models, and Poisson regression models. These statistical models are widely used in various practical fields. The models are constructed based on data set. It is an important problem that the models are evaluated appropriately by using a goodness-of-fit measure. There do not exist widely used goodness-of-fit measures such that the measure can evaluate the generalized linear models. We have clarified the statistical properties of the proposed model evaluation measures.

Free Research Field

統計学

Academic Significance and Societal Importance of the Research Achievements

世の中の生活の中には様々な統計モデルが潜んでいる。インターネットで買い物をする際、お勧めの商品をWEBサイトが推薦してくれる。これらは統計的なモデルが使われていることがあり、その良し悪しを評価する尺度を開発することは学術的な意味だけではなく、より便利な社会に役立てる一般的な社会的な意義を持ち合わせている。