How to Lie with Data Visualization

http://data.heapanalytics.com/how-to-lie-with-data-visualization/

by matm • 12 years ago

120 17 12 years ago

data.heapanalytics.com

Great article, I've seen those techniques being used many times by brazilian media. It's not easy to spot those lies at a glance but it's a must if you don't want to get fooled by traditional media news.

vitorbaptistaa• 12 years ago

(for Portuguese-speakers) I did a talk with the same title last year where I showed a few more examples in http://vitorbaptista.com/como-mentir-usando-visualizacao-de-...

These are only the slides, which lack lots of explanations, but might be interesting to someone.

Mithaldu• 12 years ago

The last example is especially striking to me, as i've seen that graph used in a discussion about the effects of gun laws. The person posting it had taken the time to remove the numbers.

notahacker• 12 years ago

I was going to say it was the most heinous deliberate visual distortion of data to create the wrong impression I'd ever seen, and then I read the BI article that was clearly slanted the opposite way from the apparent bias in the graph, and noted that they claimed the graph was originally from Reuters, an organization relatively unlikely to sacrifice their credibility on the altar of defending Florida's "stand your ground" laws. If people choosing the visualizations didn't read too much into the message they conveyed, what does that say about the average reader taking a quick glance ?

If I were to chart people that interpret visualizations to draw conclusions the narrative doesn't always express and people that see visualizations as a pretty picture to go with what the words say, I have a feeling the former group would be so small I could get away with using a pie chart[1]

[1]But seriously, emphasizing the difference between a large majority and a tiny minority is the acceptable use case for a pie chart, isn't it?

TazeTSchnitzel• 12 years ago

I suspect the flipped y axis was for sheer aesthetics - blood dripping down.

nathancahill• 12 years ago

I highly recommend reading "How to Lie with Statistics"[0]. Old but still very relevant.

[0] http://www.amazon.com/How-Lie-Statistics-Darrell-Huff/dp/039...

bashcoder• 12 years ago

Haha - we posted the same link, so I will +1 yours instead. This article is very much in keeping with the spirit of that book. A friend kindly gave me a first edition from 1954.

sillysaurus3• 12 years ago

I came here to say the same thing. Also http://www.statisticsdonewrong.com/

hvass• 12 years ago

I highly recommend it as well! There are amazing examples in the book. And if you look around, you should be able to find presentations on it online. (filetype:pdf / ppt in Google).

MarcScott• 12 years ago

This reminds me of an episode of The Saturday Night Armistice from way back in 1995

https://www.youtube.com/watch?feature=player_detailpage&v=68...

j2kun• 12 years ago

It's interesting to see the y-axis scaling example, because the author uses it to sell as incomplete a truth as the document he's criticizing.

displaying the data with a zero-baseline y-axis tells a more accurate picture, where interest rates are staying static.

This is not true of all data, or even interest rates. Who's to say that a 0.01 increase in interest rate doesn't translate to large differences in money? If this is true (and it depends a lot on the intended audience for the chart), then representing it with a zero baseline is far more deceptive and the "bad" chart does a better job showing that the small increases are in fact getting steadily bigger.

The problem is that data alone is not sufficient to say anything, and the context (in this case, noise in interest rate changes) is too important to give a single rule for your scale. How you interpret context, of course, depends on what you're selling.

001sky• 12 years ago

yep, traders measure basis points...and do big numbers x small margins, and measure cumulative profits.

aperrien• 12 years ago

It can also depend on what is the easiest thing for you to hear. Take a look at this article: http://blogs.reuters.com/great-debate/2014/04/14/america-the...

It's disturbing how much we look to confirm out preset biases.

jmharvey• 12 years ago

The article doesn't even mention the most egregious offender on display in their examples: the tilted/3-D pie chart. In the Romney/Palin/Huckabee graphic, Palin's wedge has the smallest area, despite having the largest number.

etler• 12 years ago

That's a very good point, as it's commonly used but rarely pointed out. I feel the examples in this article are very well known, but I never thought about the impact of 3D charts. I suspect many people who use it do so without any malicious intent, but just because they think it looks cool.

dsugarman• 12 years ago

a logarithmic or exponential scale could be effective in lying as well

ASneakyFox• 12 years ago

What about when people make charts where the x and y axises are unrelated or stupid? If you plot oxygen consumption and death you'd find that almost all people who died last year consumed oxygen.

How to Lie with Data Visualization

Related Stories