| |
A central element in the design and generalization of a quantitative map that partitions the range of data values into discrete categories, each assigned a unique symbol. Typically a map key links the class intervals to their respective symbols, which may vary in size, graytone value or colour. Especially common on choropleth maps, which portray data for census tracts and other areal units, class intervals are also used for maps of linear and point phenomena. Although not always apparent in a map key, class intervals are embedded in contour and other maps on which isolines divide the data into categories or layers (Jenks, 1963).
The number of categories and their divisions can have a decisive influence on the mapped pattern (Monmonier, 1996, pp. 40-3; see figure). For example, a map of mortality or poverty can overstate a threat if its classification assigns many places to the highest, most ominous category. Similarly, another map of the same data might present a markedly more optimistic view by placing far fewer areas in the upper category. Moreover, different sets of class intervals suggest or deny similarity to another distribution. Because the map author can influence interpretation by manipulating a map\'s class intervals, wary viewers might question the author\'s motives.
Deliberate bias is no less worrisome than ignorance. All too often users of desktop mapping software accept blindly an automatic, \'default\' classification that divides the range into five equal intervals. Easily programmed for a computer, an equal-intervals classification is no better or more natural than any other standardized scheme. Moreover, if the data contain one or more extraordinarily high or low values called outliers (cf. residual), an equal-intervals classification might yield empty categories or assign almost all places to a single category. A less outrageous default is the quantile scheme, which assigns an approximately equal number of places to each category. Although a five-category map identifying places in the upper and lower fifths can be meaningful to some viewers, the resulting regions are often less homogeneous than the map suggests (Evans, 1977).
Because a choropleth map is a regionalization based on a single variable, map authors often examine a univariate scatterplot, or number line, for \'natural breaks\' (Jenks and Caspall, 1971). Although readily apparent homogeneous categories merit consideration, map viewers might prefer inherently meaningful breaks such as zero or the variable\'s national or world average. On a map of population growth rates, for instance, a category break at zero differentiates places that gained from those that lost, and a category break at the national average affords a convenient assessment of relative growth.
Conscientious map authors have several options: provide viewers with a univariate scatterplot that relates class intervals to the distribution of data values; use and identify inherently meaningful breaks; or explore the effects of various class-intervals schemes and show whatever radically different alternative views might emerge. Better still is a multimedia visualization system in which a dynamic map presents multiple regionalizations and invites the viewer to verify their stability (Egbert and Slocum, 1992). (MM)
{img src=show_image.php?name=bkhumgeofig11.gif }
class interval Three different sets of category breaks (lower left) yield different map patterns
References Egbert, S.L. and Slocum, T.A. 1992: exploremap: an exploration system for choropleth maps. Annals of the Association of American Geographers 82: 275-88. Evans, I.S. 1977: The selection of class intervals. Transactions, Institute of British Geographers NS 2: 98-124. Jenks, G.F. 1963: Generalization in statistical mapping. Annals of the Association of American Geographers 53: 15-26. Jenks, G.F. and Caspall, F.C. 1971. Error on choropleth maps: definition, measurement, reduction. Annals of the Association of American Geographers 61: 218-44. Monmonier, M. 1996: How to lie with maps, 2nd edn. Chicago: University of Chicago Press. |
|
