Development of imputation methods for spatially dependent missing data

2016 Fiscal Year Final Research Report

• Project/Area Number: 26820217
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Civil engineering project/Traffic engineering
• Research Institution: Kobe University (2016); Hiroshima University (2014-2015)
• Principal Investigator: SEYA HAJIME 神戸大学, 工学研究科, 准教授 (20584296)
• Co-Investigator(Renkei-kenkyūsha): TSUTSUMI Morito 筑波大学, システム情報系社会工学域, 教授 (70292886)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: 地理情報 / 空間相関 / 欠損データ

Outline of Final Research Achievements

We developed new imputation method for spatially dependent missing data. The method attempts to apply spatial information to the prior distribution of missing values. We found that [1] By using spatial information for prior distribution, the imputation accuracy of some spatially auto-correlated explanatory variables can be improved (e.g., rent price, accessibility index). [2] Prior spatial information on missing data may not be effective for variables with a high {nugget/sill} ratio; in other words, the measurement error and/or micro-scale variation is dominant in total variation (e.g., FAR(Ratio) and age).

Free Research Field

土木計画学