Common Data Visualization Mistakes Part 4: Bar Graphs with No Zero

This video is part of the How to Avoid Common Data Visualization Mistakes series, presented by Naomi B. Robbins, Data Visualization Expert at NBR.

Transcript:

Here we have a Federal Drug Control Budget in billions. If we look, the left most bar is roughly $11 billion. The right most bar is under $20, and yet it looks several times that of $11. It's just a lie.

Here is Weather Underground saying their forecast is smarter, but I wrote, but your graph is dumber. This is how accurate forecasts are. My forecast is 70% and Weather Underground is 77%. It looks many, many times. From 70 to 77 looks like a large multiple. It's just a lie.

As I say, if you judge a graph by position, for example, a line graph, you judge by position along a scale. I say you don't need zero. But if you're judging length, you do need zero.

Break lines don't help. Here they put a zero, but then told us there's a break. It's just like the one I showed you before. If you draw something straight through the break, the break doesn't happen. So I claim that this is a lie. It makes 51 look more than double 47. And it's the same problem with this 3D, you don't know which grid line to use to read it.

This shows the number of outpatients at VA hospitals and the number of full time equivalent employees. If we look, we see bar graphs with no zero, which I hope you never realize are a lie. And we see that it gives the impression that the number of outpatients is growing dramatically, while the number of employees is fairly constant. How can these poor employees take care of so many more people?

Well, let's give it a scale of zero, since it's a bar graph, and you can see, yes, the number of outpatients are growing, but nothing like the dramatic increase that was suggested before.

And now, let's show the growth of visits and the employees, the percentage over time, and the blue is slightly above the red, but you can see what a lie the others were.