Scholarlink

<p>The spatial beta regression model is commonly used for geostatistical data defined on bounded support, primarily focusing on modeling the conditional mean of the response. However, this conditional mean model can be sensitive to skewness and outliers. In this paper, we propose robust quantile regression models for bounded spatial data, with particular emphasis on two parametric median-based hierarchical spatial regression models. Specifically, these models resemble spatial generalized linear mixed models (SGLMMs), wherein the response variable is modeled using Kumaraswamy and Johnson-t distributions. The proposed models are more robust than the usual spatial beta regression model against the presence of asymmetries and extreme observations. A fully Bayesian analysis is employed to make inferences using Markov chain Monte Carlo (MCMC) sampling algorithms. The proposed methodology is illustrated by some simulation experiments and by applying it to forest canopy cover data modelling. The results demonstrate that our proposals noticeably outperform the competitor in terms of predictive accuracy and precision of estimates.</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

Robust modeling for continuous bounded spatial data

  • 期刊论文

  • 发表日期：

    2025年2月8日

查看全文
  • Alireza Ahmadi,
  • Majid Jafari Khaledi,
  • Hormoz Sohrabi,
  • Esmaeil Najafi

摘要

The spatial beta regression model is commonly used for geostatistical data defined on bounded support, primarily focusing on modeling the conditional mean of the response. However, this conditional mean model can be sensitive to skewness and outliers. In this paper, we propose robust quantile regression models for bounded spatial data, with particular emphasis on two parametric median-based hierarchical spatial regression models. Specifically, these models resemble spatial generalized linear mixed models (SGLMMs), wherein the response variable is modeled using Kumaraswamy and Johnson-t distributions. The proposed models are more robust than the usual spatial beta regression model against the presence of asymmetries and extreme observations. A fully Bayesian analysis is employed to make inferences using Markov chain Monte Carlo (MCMC) sampling algorithms. The proposed methodology is illustrated by some simulation experiments and by applying it to forest canopy cover data modelling. The results demonstrate that our proposals noticeably outperform the competitor in terms of predictive accuracy and precision of estimates.