Perception of clusters in statistical maps

Abstract

Two experiments observed performance on a cluster identification task across a variety of common statistical maps. Stimulus maps displayed mortality rates for several diseases and subjects had to identify regions of the map that were perceived to form a cluster of particularly high (or low) mortality. Subjects marked the perceived centroid of each cluster, and analyses focused on the dispersion of centroid location across subjects. Under these circumstances, monochrome classed choropleth maps were found to minimize dispersion, compared to a two opposing colours scheme, a dot density map, a pie map, and a categorical (hue-based) colour scheme. Maps using a familiar geographical unit (i.e. a US. state) supported better recall of the information than maps using less familiar and smaller geographical units. The results were found to be interpretable within current cognitive theory.

Interpreting Spatial Patterns: An Inquiry Into Formal and Cognitive Aspects of Tobler's First Law of Geography

“…There are several studies in geography and cartography that are concerned with the perception and identification of individual clusters in maps (Slocum 1983;Lewandowsky et al 1993;Sadahiro 1997), the measure of map complexity (Olson 1975;Bregt and Wopereis 1990), similarity comparisons (Steinke and Lloyd 1981), or influences of different visualization methods on cluster detection (Walter 1993). In contrast to these studies, our research focuses on clustering (as identified by spatial autocorrelation methods) as a global phenomenon in spatial patterns and does not aim to identify the best visualization method to represent clusters.…”

Choosing Geographic Units for Choropleth Rate Maps, with an Emphasis on Public Health Applications

“…Areas can also be represented symbolically, such as by placing a color-coded or proportional point symbol at each area centroid or by placing in each area a number of point symbols proportional to the mapped statistic (a dot density map). This approach tends to result in widely spaced points in low-population areas, making pattern detection difficult (Lewandowsky et al 1993). Geographic areas can also be distorted so that the areas themselves relate to a variable such as population, but so-called cartograms or density-equalizing map projections can be problematic to construct and interpret (Dorling 1993;Merrill et al 1996).…”

Mapping Mortality: Evaluating Color Schemes for Choropleth Maps

“…In particular, there has been uncertainty about the impact of color-scheme choices on map-reading tasks typical of geographical analyses (Brewer 1994a;Lewandowsky and Behrens 1995). Among the issues we address here are the development of effective hue combinations for use in diverging color schemes, the impact of diverging schemes relative to sequential schemes, and whether or not color is necessary at all for such maps.…”

“…While maps and other visualization tools have been successful in prompting new insight about spatial pattern (for example, identification of disease clusters), map designers have been concerned that Type I and II visualization errors (seeing patterns that do not exist or missing patterns that do) are more frequent than necessary because of limited understanding of the impact of various symbolization choices on map reading (resulting in choice of inappropriate symbols) (MacEachren 1995). In particular, there has been uncertainty about the impact of color-scheme choices on map-reading tasks typical of geographical analyses (Brewer 1994a;Lewandowsky and Behrens 1995). Among the issues we address here are the development of effective hue combinations for use in diverging color schemes, the impact of diverging schemes relative to sequential schemes, and whether or not color is necessary at all for such maps.…”