Detection of Sample Differences from Dot Plot Displays

  • Authors:

  • Lisa A. Best;Laurence D. Smith;D. Alan Stubbs

  • Affiliations:

  • Department of Psychology, University of New Brunswick, Saint John, Canada NB E2L 4L5;Department of Psychology, University of Maine, Orono, USA ME 04469-5782;Department of Psychology, University of Maine, Orono, USA ME 04469-5782

  • Venue:
  • Diagrams '08 Proceedings of the 5th international conference on Diagrammatic Representation and Inference
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Cleveland and McGill [10] concluded that dot plots are effective when one judges position along a common scale. We assessed the ability of graph readers to detect sample mean differences in multipanel dot plots. In Experiment 1, plots containing vertically arranged panels with different sample sizes and levels of variability were presented. Sensitivity was greater with large samples and low variability. In Experiment 2, sensitivity depended on the location of the comparison sample, with vertical and superimposed arrays yielding greater sensitivity than horizontal or diagonal arrays. Horizontal arrays also produced a bias to judge data in right-most panels as having higher means. Experiment 3 showed that ordering of data had little effect on sensitivity or bias. The results suggest that good graph design requires attention to how the specific features of a graphical format influence perceptual judgments of data