Regression-based scaling to facilitate the cross-correlation of graduated circle maps

Most cartographic research on the communication effectiveness of graduated circles has been directed toward value estimation tasks. Apparent value rescaling of graduated circles has no inherent effect on visual cross-correlation, and another type of design modification is needed to improve judgments of similarity. Regression-based scaling is grounded in the linear regression model underlying the product-moment correlation coefficient and is a means for obtaining similar map patterns for highly correlated geographic distributions and dissimilar patterns for uncorrelated distributions. Although this design modification destroys the symbols’ ratio scale, value estimation tasks are not obstructed if legends contain representative ranges of circles sizes.