Sometimes charts and graphs are designed in a way that is very misleading. Here are some examples, and explanations for what makes them misleading.
This is a pretty famous bad graph that I have edited to remove the actual subject matter and protect the innocent. If you take a look at it, you can see that there is a point where something changed. What happened after the change?
If you said that the rate *dropped* you’re wrong. This chart was designed to give you that impression and make you think the exact opposite of the truth.
Look closely and you’ll see the vertical access is flipped, making the chart read, at a glance, the opposite of what you might think. It looks like the rate *dropped* after the change because we are used to seeing graphs and charts with a vertical axis that starts low at the bottom and increases going up.. But because the vertical axis is flipped, the truth is that the numbers sharply INCREASED after the change.
A quick look at the bar chart might make you think that the green one happens at twice the rate of the blue. But the two are actually very close!
In designing this graph, the vertical axis leaves out everything below 75. This makes 76 and 77.2 appear to be very different, when in fact they are very similar. This is the kind of thing that a researcher might do to make small differences they found *appear* to be bigger differences so they can draw more attention to their research and results.
It’s a good idea to look at the actual numbers and (once again) closely inspect the axes of the chart so you know what you’re seeing.
Do a little quick math on this pie chart. Does it add up?
While it is possible for some data on percentages to not add up to 100%, when this happens a pie chart is a lousy way to present the data. One example is if you asked a group of people where they saw your marketing materials, and had them check all that apply. In this example, maybe 43% of people clicked the box that they had seen your web site, 38% saw your Instagram account, and 24% heard a personal recommendation from a friend. It’s entirely possible some people had more than one of these be true.
But even thought this kind of information shouldn’t be presented this way, it happens all the time. And it means that some things look bigger than they are!
There’s a whole style of graph called a “bubble chart” where – in theory – the size of the bubble indicated the size of the data.
I made an example here that could be a chart showing the cesarean rates of various hospitals in the area. If I made the bubbles sized for anything other than the cesarean rates, I could make it look like on hospital had a higher than actual rate. Maybe I made the size of the bubbles reflect the number of births there annually, and then put the cesarean rate in the bubble. The chart could easily mislead!
Any time you see a chart where the size of an object is supposed to be meaningful, double check the actual numbers.
We associate certain things with color. Red is bad, green is good. Orange or yellow are hazard and caution. If you looked at this map of the US, which states would you immediately think had the worst health outcomes? This example (which I made just for this article) intentionally has no legend, but even if it did, many people wouldn’t bother to look at the legend, and would just look to see if their state ranked best, worst or other.
Missing legends are a fairly common problem. I’ve done it myself (and I don’t just mean this example chart…) Had to pull and fix a chart I’d created about holiday inductions.
These certainly are not the only problems out there, and not every bad data visualization is an intentionally misleading one. But being careful about how you look at charts can help you better grasp the concepts and not be misled by craziness. If you’d like to see more real life examples of bad data visualization, follow wtfviz on Twitter