Density in scatterplots and the estimation of correlation

This study examines the perception of bivariate normal data presented in a scatter diagram, and creates a model that successfully explains how individuals perceive the information contained in scatterplots.

Abstract

Abstract The construction of a graphical presentation involves the representation of information by means of visual symbols. The acquisition of information from the resultant graph is a perceptual process that involves the decoding and interpretation of the visual symbols. Hence good design decisions will be based on an understanding of the information acquisition process and in particular graphical perception. This study examines the perception of bivariate normal data presented in a scatter diagram, and creates a model that successfully explains how individuals perceive the information contained in scatterplots Subjects were shown a series of scatter diagrams on the CRT of a microcomputer and were asked to estimate correlation. Several variables were examined that explain estimated correlation including regression slope, dispersion, number of points displayed, and the size of the CRT screen. All of these factors were found to significantly affect subjects’ estimates of correlation.