Classification for visualizing data: integrating multiple attributes and space for choropleth display

It is common for researchers in the social sciences to be concerned with the distributional aspects of social phenomena – such as rates of unemployment, levels of household income and types of housing tenure – across spatial units that comprise a city, state or nation, and to seek to visualize variations in the patterns of such socio-spatial data in the form of a map. Commonly that involves classifying data to produce a choropleth map. In this chapter we review a number of classification approaches that are commonly used – especially by geographers – to generate map displays of socio-spatial datasets at point and polygon levels. We also discuss the development of a new methodology and tool for enhancing classification through the categorization of data to produce an improved capacity for choropleth display of data in a map. The chapter first discusses standard categorization routines such as equal interval, quantile and natural breaks, and the Location Quotient (LQ) which is a benchmarked approach to categorization. Performances among the various classification approaches may be compared by considering the total within-group variance (TWGV) and the total within-group difference (TWGD), the measure structured in the median clustering objective.