ABSTRACT
Analysis of rare events (e.g., cancer, AIDS, homicide) often suffers from the small population (numbers) problem, which can lead to unreliable rate estimates, sensitivity to missing data and other data errors (Wang and O’Brien, 2005), and data suppression in sparsely populated areas. The spatial smoothing techniques such as the floating catchment area method and the empirical Bayesian smoothing method, as discussed in Chapter 3, can be used to mitigate the problem. This chapter introduces a more advanced approach, namely, “regionalization,” to this issue. Regionalization is to group a large number of small units into a relatively small number of regions while optimizing a given objective function and satisfying certain constraints. The chapter begins with an illustration of the small population problem and a brief survey of various approaches in Section 9.1. Section 9.2 discusses two GIS-based approaches, the spatial order method and the Modified Scale-Space Clustering (MSSC) method. Section 9.3 explains the foundation of Regionalization with Dynamically Constrained Agglomerative Clustering and Partitioning (REDCAP) method, and Section 9.4 uses a case study of analyzing late-stage breast cancers in the Chicago region to illustrate its implementation. The chapter concludes in Section 9.5 with a brief summary.
