Bar Chart Axis Musings
Researching data best practices, I’ve heard it said that you should always start a bar graph y-axis at 0 because not doing so will irreconcilably skew your data and introduce bias. I accept that that is usually true but what about situations in which it would be better to start at another number? My contrarian nature demanded that I investigate, and I came up with this situation.
What if we’re measuring body temperature for fifteen individuals? We care about temperatures that are too high and also too low. Maybe it’s to show the results of a medication test. Maybe they are a foraging club (but not very smart), and they all got nervous after eating strange mushrooms. Whatever the situation, we have a small chart with everyone’s temperature and a constant line marking 98.6.
We can see that all of the values cluster around the constant, but that’s going to be the case for any living person.
Meanwhile, if I change the y-axis range to start at 90, I get a much clearer picture.
领英推荐
Jasmin looks like she has a bad fever while Tim might be suffering from hypothermia. With a variable like this, minor differences make a big deal. When your y-axis range is the body temperature for a living human, it’s physically impossible for it to fall below 56 and extremely and very unlikely for it to fall below 80. If a zero is not possible, does it make sense to include it in the chart?
It would be a different situation if we’re measuring something else where a zero is possible, like grades. If these charts are meant to show a final exam score, the first chart is more representative. It wouldn’t really matter that Jasmin did some extra credit and got a 102 or that Tim messed up a few multiple-choice questions and got a 95; all of these students’ parents would be pleased with these test scores.
So, I think I found a flaw in this directive, but what do we do here? The obvious solution is not to make bar charts of body temperature if it causes so much headache. But is there anything more? Did I find not just one example but a series of cases that breaks this rule? Can anyone think of other examples in which a range of data in a bar chart cannot possibly go to zero? How would that look?
?Data Analytics Expert ?Internat'l Keynote Speaker ?CTO? ??? ?Amazon Best-selling Author??Senior Systems Specialist ??? ?? ?Gartner Peer Community Ambassador of the Year 2023
5 个月You've raised an excellent point here, Leah! The "rule" to start a bar chart’s y-axis at zero is simply to avoid misleading viewers, but your example with body temperatures shows an exception where it makes total sense to adjust the axis. If zero isn't practical or relevant, like in this case, adjusting the range can provide clarity -- perfectly fine to do so. However, it's crucial to be transparent about these adjustments to prevent misinterpretation. A small note or annotation can help. This balance allows for better visual storytelling without compromising the integrity of the data. Thank you for mentioning me, Leah; I really appreciate your valuable insights, ma'am!