Spatial regression modeling estimating a wide variety of effects from diverse data

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K13261
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 04010:Geography-related
• Research Institution: The Institute of Statistical Mathematics
• Principal Investigator: Murakami Daisuke 統計数理研究所, 統計基盤数理研究系, 准教授 (20738249)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 時空間回帰 / 非ガウスデータ / 高速化 / spmoran

Outline of Final Research Achievements

This study developped spatio-temporal regression models for estimating (a) diverse effects from (b) diverse data, and implemented the developped methods in packages in a free statistical software R. Regarding (a), spatial regression was extended to allow automatic estimation of data distributions by introducing a compositional transformation function. It was confirmed that the developped method flexibly handle a wide range of data distributions, including Box-Cox and Tukey g-and-h distributions. Regarding (b), we developed a new spatio-temporal model that can handle spatial, temporal, and spatio-temporal correlations on multiple time axes simultaneously. The usefulness of the methods developed in (a) and (b) was confirmed by applying them to a wide range of real-world data, including residential land prices, the number of crimes, and the number of positive Covid-19 cases.

Free Research Field

空間統計学

Academic Significance and Societal Importance of the Research Achievements

既存の空間統計手法の多くは、ガウス過程に依拠したものであり結果的に計算コストや柔軟性に課題が残されていた。本研究では、それらの課題に対処して、計算コストを維持しながら幅広いデータと効果を扱えるように空間統計手法を拡張するものであり、同分野の発展に寄与する学術的意義の大きい研究である。また開発手法をフリーの統計ソフトウェアRのパッケージを通して公開しており、地理情報に関する実務者・研究者を、手法提供の観点で支える社会的意義の大きな研究である。