Chris

Lee

Content

Copywriting

Design

Inconsistent Scales Lead to Flawed Conclusions


November 19, 2019

The graph above shows obesity rates in the US in the year 2000, with darker colours meaning more obesity.

Here’s the graph showing the same, but for 2018:

It doesn’t look like much has changed, right? Only four more states report higher obesity, indeed three report less. A net gain of one state that is more obese than in 2000.

(The black circles show states that reported obesity in 2000 but not in 2018. The red circles show the opposite.)

From this finding, it’s hard to believe that America is facing a health crisis.

But let’s take a look at the legends that came with the original maps on the CDC website, where the raw data was taken from. 2000 on the left, 2018 on the right:

They’re completely different. For a state to be dark blue in 2000, obesity rates had to be above 21.8. This means that all dark blue states in the 2000 map would be light green if we used the 2018 legend.

If we make the legends and colouring consistent between the maps, here’s what we see for 2018:

When using the data ranges from 2000, every single US state is in the top bracket for obesity. Over 21.8% of adults in every state report as obese.

From the first two graphs we saw, you’d be forgiven for thinking things weren’t too bad:

But the actual picture is a lot more severe:

Using consistent scales when presenting different datasets is vital if you want to make it possible to draw clear conclusions from the data.

Which we all do, right?

Otherwise, why would you include graphs at all..?