Geographically Optimal Similarity

Abstract Understanding geographical characteristics of distribution patterns and spatial association is essential for statistical inference such as factor exploration prediction. The similarity principle was recently developed to explain the between variables. It describes comprehensive degree approximation a structure instead explicit relationships However, there are still challenges similarity-based methods. For instance, all samples used prediction, leading increased calculation burden reduced prediction accuracy due noise unrelated data in large sets. This study develops geographically optimal (GOS) model accurate reliable based on principle. In GOS, configurations first characterized, similarities unknown known observation locations assessed. Next, an threshold determined select small number observations with at each location. Finally, uncertainty assessment approach assess map uncertainties GOS predictions. A new R package “geosimilarity” conduct models. this study, implemented predicting distributions trace elements mining region Australia. Results show that can use derive more predictions than linear regression basic configuration addition, pattern be improved by eliminating phenomenon wherein clustered near mean values contain striped textures. Therefore, demonstrates greater potential implementing bringing information from relatively high any location across space effective broader fields practice.