A Classification Method for Choropleth Maps Incorporating Data Reliability Information

Observations assigned to any two classes in a choropleth map are expected to have attribute values that are different. Their values might not be statistically different, however, if the data are gathered from surveys, such as the American Community Survey, in which estimates have sampling error. This article presents an approach to determine class breaks using the class separability criterion, which refers to the levels of certainty that values in different classes are statistically different from each other. Our procedure determines class breaks that offer the highest levels of separability given the desired number of classes. The separability levels of all class breaks are included in a legend design to show the statistical likelihood that values on two sides of each class break are different. The legend and the associated separability information offer map readers crucial information about the reliability of the spatial patterns that could result from the chosen classification method.