The Nature of Geographic Information

18. Two Classification Schemes

PrintPrint

Many different systematic classification schemes have been developed. Some produce "optimal" classes for unique data sets, maximizing the difference between classes and minimizing differences within classes. Since optimizing schemes produce unique solutions, however, they are not the best choice when several maps need to be compared. For this, data classification schemes that treat every data set alike are preferred.

Screenshot of the ArcMap classification window
Figure 3.19.1 Portion of the ArcMap classification dialog box highlighting the schemes supported in ArcMap 10.2.

Two commonly used schemes are quantiles and equal intervals ("quartiles," "quintiles," and "percentiles" are instances of quantile classifications that group data into four, five, and 100 classes respectively). The following two graphs illustrate the differences.

Graph showing county percent population change
divided into five quantile categories
Figure 3.19.2 County population change rates divided into five quantile categories.

The graph in Figure 3.19.2 groups the Pennsylvania county population change data into five classes, each of which contains the same number of counties (in this case, approximately 20 percent of the total in each). The quantiles scheme accomplishes this by varying the width, or range, of each class.

Graph showing county percent population change divided into
five equal interval categories
Figure 3.19.3 County population change rates divided into five equal interval categories.

In the second graph, Figure 3.19.3, the width or range of each class is equivalent (8 percentage points). Consequently, the number of counties in each equal interval class varies.

PA map showing the quantile classifications of the percent
population changes for each county
Figure 3.19.4 The five quantile classes mapped.
PA map showing the equal interval classifications of the
percent population changes for each county
Figure 3.19.5 The five equal interval classes mapped.

As you can see, the effect of the two different classification schemes on the appearance of the two choropleth maps above is dramatic. The quantiles scheme is often preferred because it prevents the clumping of observations into a few categories shown in the equal intervals map. Conversely, the equal interval map reveals two outlier counties which are obscured in the quantiles map. A good point to take from this little experiment is that it is often useful to compare the maps produced by several different map classifications. Patterns that persist through changes in classification scheme are likely to be more conclusive evidence than patterns that shift.