Bad Charts & Bias

Bad Charts & Bias

I am always on the look out for misleading or biassed use of data.?This graphical example, which showed up in my Twitter feed on Friday, is a particularly egregious one.

No alt text provided for this image

Edward Tufte, in his excellent 2001 book “The Visual Display of Quantitative Information” in the chapter titled “Graphical Integrity”, defines the Lie Factor of a chart as:

Lie Factor = size of the effect shown in the graphic / size of the effect in the data

If we assume the data labels are correct, the Lie Factor for this chart, considering the 2021 versus 2018 bars which both have values of 336 is 1.97.?Anything outside of the range 0.95 - 1.05 Tufte defines as substantial distortion.?Distortion of this scale should have triggered alarm bells before this chart made it to air.

Another issue with the chart is the colours.?They are a meaningless distraction, yet viewers often assign meaning to colour which can further confuse or mislead. Tufte refers to this as Chartjunk. Lastly, when I am reviewing data and charts in my day job, I am suspicious of repeated values such as we have here from 2018 to 2021. ?

Here’s what the chart should look like, if the values on the bars can be believed.?

No alt text provided for this image

Unfortunately it transpires that the bar heights in the original are correct but the data labels for 2019-2021 are wrong.?It’s vital that all data is presented in clear and unbiassed ways, and that there are appropriate quality control measures in place to trap and fix errors.?That goes double when it is data on such a sensitive topic being shown to a large audience.

Love it - will be discussing the lie factor today :)

Rob Urquhart

Global Lead, Consulting/SI Partnerships at Meta

2 年

User error! ????♂?

That is a corker of a chart..(not) we must have a conversation about Tufte & story telling with data…

Gareth Mitchell-Jones

Enabling CXOs, their teams, and organisations to deliver impact, value and outcomes with data and technology since 2002. I work to increase revenues, reduce costs & mitigate risks from strategy into operation.

2 年

What - no QA of the visualisations? m 00 m Based on the scale alone one could say that 2021 must be 570-628, 2020 2019 385-410 490-570

要查看或添加评论,请登录

Conor McGovern的更多文章

社区洞察

其他会员也浏览了