Graph Construction

Abstract

Graphs are an effective and compelling way to present scientific results.  With few rigid guidelines, researchers have many degrees-of-freedom regarding graph construction.  One such choice is the range of the y-axis.  A range set just beyond the data will bias readers to see all effects as big. Conversely, a range set to the full range of options will bias readers to see all effects as small.  Researchers should maximize congruence between visual size of an effect and the actual size of the effect.  In the experiments presented here, participants viewed graphs with the y-axis set to the minimum range required for all the data to be visible, the full range from 0 to 100, and a range of approximately 1.5 standard deviations.  The results showed that participants’ sensitivity to the effect depicted in the graph was better when the y-axis range was between one to two standard deviations than with either the minimum range or the full range.  In addition, bias was also smaller with the standardized axis range than the minimum or full axis ranges.  To achieve congruency in scientific fields for which effects are standardized, the y-axis range should be no less than 1 standard deviations, and aim to be at least 1.5 standard deviations.