Excel charts: how to lie or not

This is a really excellent and useful tip sheet from statistician Nick Barrowman. Really, check it out - you’ll be glad you did…

2 Responses to “Excel charts: how to lie or not”


  1. 1 Will Howard

    Yes but no. Forcing the y-axis origin to be zero, in the author’s bar-chart example is not necessarily a good, or necessarily a more honest, idea. The range of 2.5 in the density example may be significant, and if so it may be more misleading to plot the data with a y-axis origin of zero as this would visually underrepresent the variability.

    Same principle applies to the x-axis of a chart. In time-series, of course the origin and scale of the x-axis are chosen depending on what interval or process one is trying to represent. Following the suggestion of a zero-origin rule literally, a chart of the last ten years’ of any financial data set (say, the DJIA) would have to start at the year Zero A.D.?!?! In which case the chart would be unreadable, in itself a potentially misleading representation.

  2. 2 Steve Darden

    In your examples, including the origin obscures the content. In such cases, I understood the author’s advice was to use line or XY-scatter charts, instead of bar charts. Because bar charts were only suitable for presentation of “variables that are measured on ratio scales”.

    Does that make sense?

Leave a Reply






Bad Behavior has blocked 3040 access attempts in the last 7 days.