The y-must-you-insist-it-starts-at-0? axis

I keep seeing accusations that graphs where the y axis doesn’t start at 0 are misleading. Sometimes these accusations are correct. Frequently, however, these accusations are unfounded.

It’s certainly true that the relevant axis for a bar chart should include 0. But that’s not necessarily the y axis and it doesn’t necessarily mean the axis should start there; If we have positive and negative axes then 0 should appear somewhere in the middle.

Unlike bar charts, line charts and many other kinds of chart don’t encode data using length. So these charts have different rules. Yes, line charts that don’t include 0 on the y axis can be used to mislead. But so can line charts that do explicitly include 0 on the y axis, as so well demonstrated by the chart here which emphasises the supposed importance of the utterly meaningless 0 Fahrenheit in an attempt to prove a point. Unfortunately for the author, the only point it really proves is that they should leave climate science to those who actually understand temperature scales (some of the responses are excellent, however).

What about logarithmic scales? Are we meant to reject as misleading all work that uses charts with logarithmic y axes?

And why is it always the y axis people focus on? People rarely claim a time series should go back to the very start of a data series (or the Big Bang). Somehow removing empty space is seen as more of a sin than removing space that actually contains data!

It’s not just me who gets annoyed by this; if it was I’d start to wonder if I was just plain wrong. So I thought I’d collect together a list of articles expressing the same point. If you’re still not convinced by the above, try reading the articles below. If you’ve got any more suggestions for the list or would still like to tell me that I’m wrong then feel free to leave a comment.

2 thoughts on “The y-must-you-insist-it-starts-at-0? axis

  1. Willie

    Stephen Few in “Show Me The Numbers” makes a big deal about stating that bar graphs without a zero baseline are misleading. However, he does say that line graphs and dot plots can be displayed without a zero baseline, since they’re focused on change / movement, not magnitude.

    Reply
    1. Tim Brock Post author

      Indeed. Pages 106-107 and 195-197 of the second edition cover this. In Now You See It (1st ed., page 93) he also states (my emphasis): “Begin and end the scale at round numbers, and make intervals round numbers as well” in a subsection titled “Optimal Quantitative Scales”.

      Reply

Leave a Reply

Your email address will not be published. Required fields are marked *