Polygons, stars, and clusters

Abstract

One technique for displaying a set of quantitative variables is to represent the set as a polygon. Such displays allow the observer to visualize complex information quickly, as a whole. Polygon displays have been employed to display information for analysis, status, or presentation. An experimental investigation was undertaken to ascertain the effect of variation in certain visual features of the display on the consistency with which people categorize information presented as polygons. Variables included background information of the display, shading, and form. Subjects performed a categorization task on two sets of data; the results are analyzed for consistency between individuals and for consistency with certain standard clustering algorithms. The effects of distinctive portions of the figures on the judgment of similarity, and of the nature of the data and of interactions of combinations of the variables used in the experiment on the consistency of clustering were noted. Implications for the design of polygon displays are discussed.