Scholarlink

<p>In this paper, we introduces a novel partially functional linear varying coefficient geographically weighted autoregressive model for analyzing spatial data with scalar responses, functional data and scalar covariates. The proposed methodology extends conventional geographically weighted regression (GWR) by incorporating functional data analysis, thereby providing a powerful framework for investigating spatial heterogeneity in regression relationships involving functional predictors. Through the integration of functional principal component analysis (FPCA) with local linear smoothing techniques, we derive consistent estimators for three key components: (1) parametric coefficients for scalar covariates and spatial lag, (2) slope functions for functional covariates, and (3) spatially varying coefficient functions. We further establish a hypothesis testing procedure to assess the spatial stationarity of regression coefficients. Under appropriate regularity conditions, we prove the asymptotic properties of our estimators and derive their convergence rates. Extensive Monte Carlo simulations demonstrate the superior finite sample performance of our approach. Finally, we present an empirical application to meteorological data that show the practical utility of our methodology in environmental research.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Estimation and Testing of Partially Functional Linear Varying Coefficient Geographically Weighted Autoregressive Model with the Application to the Environmental Data

  • 期刊论文

  • 发表日期：

    2026年2月24日

查看全文
  • Lin Wu,
  • Yang Zhao

摘要

In this paper, we introduces a novel partially functional linear varying coefficient geographically weighted autoregressive model for analyzing spatial data with scalar responses, functional data and scalar covariates. The proposed methodology extends conventional geographically weighted regression (GWR) by incorporating functional data analysis, thereby providing a powerful framework for investigating spatial heterogeneity in regression relationships involving functional predictors. Through the integration of functional principal component analysis (FPCA) with local linear smoothing techniques, we derive consistent estimators for three key components: (1) parametric coefficients for scalar covariates and spatial lag, (2) slope functions for functional covariates, and (3) spatially varying coefficient functions. We further establish a hypothesis testing procedure to assess the spatial stationarity of regression coefficients. Under appropriate regularity conditions, we prove the asymptotic properties of our estimators and derive their convergence rates. Extensive Monte Carlo simulations demonstrate the superior finite sample performance of our approach. Finally, we present an empirical application to meteorological data that show the practical utility of our methodology in environmental research.