Visual Tools for Detecting Influential Observations in Bivariate Geostatistical Data

Abstract

This paper presents an extension of the hairplot method for detecting and visualizing influential observations in bivariate geostatistical models. To overcome the limitation of considering a single lag in semivariogram construction, we incorporate Andrews curves, allowing for a more comprehensive analysis. Additionally, we propose a novel approach that integrates boundary curves, providing a more rigorous methodology for detecting influential points. The effectiveness of the proposed methodology is assessed through simulation studies under different scenarios and disturbance levels and further demonstrated using a real soil dataset from southern Wisconsin. This application offers valuable insights into the impact of land management on carbon and nitrogen storage. By combining hairplots, cross-semivariograms, Andrews curves, and boundary curves, our approach enhances the diagnostic capabilities of spatial data analysis. This paper presents an extension of the hairplot method for detecting and visualizing influential observations in bivariate geostatistical models. To overcome the limitation of considering a single lag in semivariogram construction, we incorporate Andrews curves, allowing for a more comprehensive analysis. Additionally, we propose a novel approach that integrates boundary curves, providing a more rigorous methodology for detecting influential points. The effectiveness of the proposed methodology is assessed through simulation studies under different scenarios and disturbance levels and further demonstrated using a real soil dataset from southern Wisconsin. This application offers valuable insights into the impact of land management on carbon and nitrogen storage. By combining hairplots, cross-semivariograms, Andrews curves, and boundary curves, our approach enhances the diagnostic capabilities of spatial data analysis.