A0580
Title: A Moran eigenvector-based spatially varying quantile regression (MSVQR) model
Authors: Zhan Peng - Tohoku University (Japan) [presenting]
Ryo Inoue - Tohoku University (Japan)
Abstract: Spatial data analysis requires distinct perspectives and methodologies to understand the formation of spatial phenomena and uncover the complex interactions among factors. One of the most fundamental characteristics of spatial data is spatial heterogeneity, which means that the generation process of spatial phenomena is not always the same in space. Spatially varying coefficients (SVC) models have been widely applied to characterize spatial heterogeneity by estimating distinct values of regression coefficients at each location. In addition to the variations in space, recent studies suggest that spatial data can also exhibit variations across levels of the probability distribution. However, this issue has not been fully addressed in the analysis, which can hinder a comprehensive understanding of the underlying relationships in spatial data. To address this research gap, a Moran eigenvector-based spatially varying quantile regression (MSVQR) model is proposed. MSVQR uses Moran eigenvectors to estimate SVCs at multiple quantiles of the conditional distribution of the dependent variable. The location- and quantile-specific coefficients enable us to identify the heterogeneity across geographical space and continuous data distribution simultaneously. The proposed approach was evaluated through simulation experiments and applied to rental housing price data in Tokyo, Japan.
