White, D., and J.C. Sifneos. 2002. Regression tree cartography. J. Computational and Graphical Statistics 11(3)600-616. WED-00-014
We illustrate several types of cartographic displays that can enhance understanding from hierarchical analysis techniques such as regression trees. When the observations have spatial locations, maps of the predicted values, maps of the residuals, and maps of the predicting relationships of the tree may help to reveal associations between predictors and response. We propose an objective method for constructing maps that may help to show the geographical similarities and differences between observations based on their positions in the prediction tree. This mapping method divides the color spectrum to assign colors to the leaves, using the same hierarchical pattern as the prediction tree does to divide the data. We illustrate regression tree cartography with two examples and suggest how the prediction tree mapping method could be used for classification trees and for dendrograms produced by hierarchical clustering methods.