Ways to Lie with Charts (2014) | Hacker News Reader

Ways to Lie with Charts (2014)

nautil.us

109 pointsraldu20 comments

A long time ago I had taught a class on data visualization, and used this photo as example of creative use of pie charts:

https://www.answerminer.com/static/489716080133492e99fdcb9c8...

Here was another list: http://avoinelama.fi/hingo/kirjoituksia/misleadingvisualizat...

Those are excellent examples of the evil pie charts do. The one which i've hit most often in the world of tech company business reports is the nefarious two pie chart comparison. The very first question one should ask in that case is if the absolute aggregate (sum) is provably the same from chart A to chart B. In one notable case we finally learned that the total amounts had tripled between the two, and corresponding pie slices which had all but disappeared from A to B actually reflected quantitative growth.

What's wrong with the pie chart?

Edit: I should have read the article before the comments... Apple is using the thick wedge of foreground slice to make their share seem bigger

I think they're also lumping Android into the other category.

Not sure when that was taken but it’s totally possible that Android was legitimately a tiny portion of the market or even not yet released

Judging from the fact that RIM (the makers of the Blackberry line of devices) is shown to have 39% of the market in that chart, this must've been quite a while ago indeed.

My school spent quite a lot of time analysing charts, and the way they can be misleading. I think it was science (GCSE UK), but it was a while ago. Anyway that turned out to be a useful lesson.

I don't find your example too egregious, it has percentages at least.

Loved the 2nd list but I wish all the people who lied got handcuffs too. Don’t let them off so east!

For anyone interested in not lying with charts: https://www.edwardtufte.com/tufte/

This reminds me of one of my favorite internal wiki posts at the megacorp I work at. A particularly senior engineer (who was known for being witty and quick with puns) had written an article titled something along the lines of, “How to lie with iGraph” or maybe “lies, damn lies, and operational metrics.” I don’t quite remember the title, but it was fully of hilarious and quite specific tricks relating to our internal graphing software, and it came with a ridiculous narrative written in his characteristic style. I should find that and send it to my team tomorrow. It contains great information on how to avoid making hard to read graphs, and how to spot bad ones. It also happens to be an amazing primer on some of the more advanced features of the graphing dashboard.

I can recommend a in my opinion very interesting book about "abusing stastics" it's called "Standard Deviations: Flawed Assumptions, Tortured Data and Other Ways to Lie with Statistics". (I don't know the author or have any affiliation with it).

See also the classic book How to Lie with Statistics: https://en.wikipedia.org/wiki/How_to_Lie_with_Statistics

I highly recommend this short classic to everyone. A very fast read with excellent, timeless examples. A must-read for every high schooler too!

The swindling shape example is needlessly pushing it by swapping colours.

Reporting on stock etc markets is so rife with "Honey, I Shrunk the Scale!" that it's the norm.

And the suicide rate vs science and technology spending thing is a classic example of correlation <> causation. In this case, I suspect that they're both ~population.[0]

0) http://www.worldometers.info/world-population/us-population/

I didn't find anything particularly wrong with that graph (suicide rate v science)?

Sure if you use that data to say there is causation then that's wrong, but there isn't anything wrong with the display of data.

TFA made the point that just showing the data that way leads people to think there's causation.

And in just the same way that you shouldn't believe a article saying there's causation and just showing this graph as evidence. Then we shouldn't believe this article when they make a statement like that.

Anyway it doesn't even make sense, causation means X causes Y, the very next question is which causes which? The graph has nothing to say on that. All the graph implies is that there is a correlation, which isn't controversial (scientifically, not socially speaking).

You and I, and most HN readers, are a lot better at making those sorts of distinctions than the norm.

And correlation between science spending and suicide? I'd say that they both correlate to population, so their relationship is just confounding. But then, I'm no statistician.

They didn't mention where you start the graph.

An origin of zero and and origin of 100 make something seem suitably high or low.

And a global temperature graph starting from the late 90s will look different to one starting at the beginning of the 20th century.

Reply on news.ycombinator.com