Conditioned Choropleth Maps and Hypothesis Generation

Abstract

The article describes a recently developed template for multivariate data analysis called conditioned choropleth maps (CCmaps). This template is a two-way layout of maps designed to facilitate comparisons. The template can show the association between a dependent variable, as represented in a classed choropleth map, and two potential explanatory variables. The data-analytic objective is to promote better-directed hypothesis generation about the variation of a dependent variable. The CCmap approach does this by partitioning the data into subsets to control the variation in the dependent variable that is associated with two conditioning variables. The interactive implementation of CCmaps introduced here provides dynamically updated map panels and statistics that help in comparing the distributions of conditioned subsets. Patterns evident across subsets indicate the association of conditioning variables with the dependent variable. The patterns lead to hypothesis generation about scientific relationships behind the apparent associations. Spatial patterns evident within individual subsets lead to hypothesis generation that is often mediated by the analyst's knowledge about additional variables. Examples showing applications of the methods to health–environment interaction and biodiversity analysis are presented.