From the October 10 Boston Metro. Oops. [via]
wow…colors that have no relevance and a 200% pie chart
I don’t think it sums to 200%, either?
They forgot to include undecideds, which is why it ends up less than 200%. So instead of showing part to whole relationships, which pies/donuts are good for, this shows part to part of two separate wholes that have been merged into one for no obvious reason…
Either the colour choices make no sense (on one half, Obama’s light green, on the other half, it’s reversed), or more likely, the labels are the wrong way round on top. The “49% Gallup Obama” label points to an area that is smaller than the “46% Gallup Romney”. But even if you treat the top left segment as 46, it looks smaller than the 45 below it…
Then there’s the fact that they made the text of half of the label of the leader in each poll bold (an ugly fudge that at first just looks like clumsy accidental inconsistency), presumably after noticing that it’s virtually impossible to tell from looking at the chart (so much so that the mistake went unnoticed…). That this was necessary really, really should have been a clue to use a more appropriate chart…
It’s not as bad as it seems at first glance, but it’s not great. It’s comparing two polls – and you can see ONE of the poll-separating dividing lines.
I’d say that it’s actually worse than it seems at first glance. But that’s actually part of the brilliance of its bad-ness. I went through these phases when I looked at it: The layout is very sloppy, the arrows are inconsistently positioned, and shouldn’t be the same color as the donut pieces if they overlap them. The sums aren’t going to add up to 100%, are they? Oh – wait – it’s OK. It’s comparing two polls – and you can see one of the poll-separating dividing lines. Oh – wait, it’s worse. Neither individual poll sums to 100%. There’s no way I can tell which numbers are bad, and in which way I could attempt to mentally fix it. The colors don’t even suggest which polls the parts belong to. The donut implies parts of a whole, but this has nothing to do that. Gaah, I’m being trolled by a graphic. There’s no way this could be serious!
Well, there don’t have to be ‘bad’ numbers. Poll numbers for two options never add up to 100% because there are the ‘other’ and ‘undecided’ answers.
All other critiques are obviously valid and agreed.
A simple dot plot would handle this well – http://jsfiddle.net/jlbriggs/7tRkm/embedded/result/
Or a column if you prefer – http://jsfiddle.net/jlbriggs/7tRkm/1/embedded/result/
Ow! My intelligence!
Ugly donut for sure.
It’s wrong on so many levels. What I think it’s trying to say is that Pew’s poll results are different from Gallop’s. Because, you know, all polls should agree ;). It’s totally the wrong chart. But what do you expect from a free newspaper that’s mostly used to cover the sticky spots on the floor of the commuter rail.
I think Metro uses some templates for charts and the size of the slices or bars or bubbles is not something important to them. I saw many times the same bubble chart, that they use like a pie chart, with different numbers – but the same size of bubbles.
If they can’t afford a graphic artist, just don’t use charts.
That’s terrible. Surely there are laws against knowingly, deliberately misleading like that?
There is precisely one piece of information in that: “Conflicting polls”
I wonder if maybe spreadsheeting software will now see any column of percentages and generate a pie chart. Because that’s basically what’s happened in someone’s head here: “gotta chart up these numbers and bung them in for the print run… percentages, easy, pie chart, green green light green, pretty, done”
Even the numbers are useless without knowing the error in each number, the date of the poll. Might as well just say “Conflicting polls: numbers too close to call, but slightly different”. Come to think of it I disagree with my initial point now, the polls could well agree within the margin of error. So the one key takeaway is:
Breaking news: “Obama/Romney polls error margin approximately 3%, candidates drawn”
Become a member.
Learn to visualize your data.
From beginner to advanced.
What you get
I call myself a statistician, because, well, I’m a statistics graduate student. However, the most important things I’ve learned are less formal, but have proven extremely useful when working/playing with data.
People get married at various ages, but there are definite trends that vary across demographic groups. What do these trends look like?
The individual data points of life are much less predictable than the average. Here’s a simulation that shows you how much time is left on the clock.
There are rules—usually for specific chart types meant to be read in a specific way—that you shouldn’t break. When they are, everyone loses. This is that small handful.