Multiple Spatial Resolutions for Categorical Variables
Analysis at multiple spatial resolutions allows insight concerning the spatial relationship between False Alarms and Misses for a category. A coarsening algorithm converts fine-resolution observations into blocks that have a coarser spatial resolution, which can cause each block to have membership to more than one category. This chapter shows how to construct a square contingency table when a block can have membership to more than one category. Then the concepts of Chap. 4 analyze the contingency table to compute results at each resolution. The sum of Exchange and Shift shrinks as the spatial resolution grows coarser. The spatial resolutions at which these components shrink gives insight to the spatial allocation of a category’s False Alarms and Misses. Relevant software includes the CROSSTAB module in TerrSet available at https://clarklabs.org and the diffeR package available at https://cran.r-project.org/web/packages/diffeR/index.html.