The BBC has analyzed more Brexit referendum data; the first chart in their article caught my eye:

It seemed strange to split the x-axis at 50%. Surely this should have been split at the median of the data? (Or the mean if the data was normally distributed; but squinting at it, that doesn’t appear to be the case here.)

(They haven’t released their data, but a quick google suggests graduates are approx-25% of adult population, and that matches eye-balling the chart, so I’ll go with that.)

But that got me thinking of how to draw this chart as objectively as possible. Would it be better to keep the x-axis linear, and have the two boxes on the right be three times as wide? Or keep the boxes all the same size, and squish and stretch the x-axis, so that every N pixel span covers the same number of data points? Or something else?

**Answer**

I think this FT version of similar data serves as a decent answer about how to present the data fairly.

Rather than absolutes on a 0 – 100 scale, it zooms in to focus on the change. The lines help verify the pattern which is hard to assess on the points alone because of all the overstriking. (How many of the 1070 wards can you make out in the original?)

**Attribution***Source : Link , Question Author : Darren Cook , Answer Author : xan*