A sharp eyed statistics sleuth on reddit, skafaceXIII, spotted this little doozy from the Herald Sun on the question “Should women have to cover up when breast feeding?”:
Most people would just look at this and think “Overwhelming majority say to cover up”
Clearly you’d see all that green and think that only a small percentage differ from the masses.
But if you look at the percentage figures – 56% to 44% – that should look a lot closer than it is.
Let’s see what the figures would roughly have to be to look that way on a doughnut chart (it’s rough because I have only that picture to go off (thanks to another redditor - thetoethumb for the side by side idea). There’s another view of this at the source.
But what ARE the stats that the herald sun is using to get that bad a graph?
How the figures would have to be dodgied to look like that.
The visual representation is that of 80/20% results – or less than half the dissenting votes. So around 70 votes say “no”, 300 say “yes”. That’s a bit of a difference from 206 saying yes. About 100 different.
I think that’s pretty damned deceptive.
[...] that are misleading, badly constructed, incorrectly specified, or just plain wrong. There was a wonderful one in the Herald Sun recently, which has had considerable publicity. We hope it was just an error, and nothing more [...]