"Correlation may not imply causation, but it sure can help us insinuate it."
December 13, 2011 4:09 AM   Subscribe

Correlation or Causation? Statistics are easy: All you need are two graphs and a leading question.
posted by beaucoupkevin (29 comments total) 33 users marked this as a favorite
 
We will never know, as they absentmindedly left out the graph depicting "Correlation" and "Causation".
posted by Bokononist at 4:28 AM on December 13, 2011 [5 favorites]


Page 2, US Senate Approves War on Mountain Range.
posted by Wolfdog at 4:33 AM on December 13, 2011 [2 favorites]


"Is this mountain range affecting the murder rate?"

I love that one so very hard. Some random chartline, scaled to match the reality of the geographic landscape of...where, exactly? It must be so, peaks don't lie.

Brilliant.
posted by iamkimiam at 4:35 AM on December 13, 2011 [4 favorites]


Also, that mountains have agency. To kill New Yorkers.
posted by iamkimiam at 4:36 AM on December 13, 2011


Nice, thanks!
posted by carter at 4:46 AM on December 13, 2011


I like this very much.
posted by OmieWise at 5:13 AM on December 13, 2011


"Is this mountain range affecting the murder rate?"

Reminds me of Demetri Martin's "ability to draw mountains over time" graph
posted by kersplunk at 5:33 AM on December 13, 2011 [1 favorite]


Speaking of deceptive graphs, I liked this.
posted by benito.strauss at 5:41 AM on December 13, 2011 [10 favorites]


Damm benito.strauss, there is no depths they won't sink to!
posted by CautionToTheWind at 5:45 AM on December 13, 2011


This mountain range looks suspicious. I think we don't see the whole picture - the actual number of murders might be way bigger!
posted by hat_eater at 5:57 AM on December 13, 2011 [1 favorite]




Good thing nobody actually uses such laughably ridiculous tactics! Because the real way to go is to take a graph that shows something you disagree with, rotate it by 45 degrees and then claim that there is no trend! (Page 18)
posted by Pyrogenesis at 6:42 AM on December 13, 2011


My age; inflation.

DO THE MATHS, PEOPLE.
posted by clvrmnky at 6:57 AM on December 13, 2011


the real way to go is to take a graph that shows something you disagree with, rotate it by 45 degrees...

I was baffled until I got to 'Christopher Monckton sensitively extracted the introductory main points, redrew several graphs...' The number one rule of drawing graphs is surely 'don't ask Monckton'.
posted by robself at 7:03 AM on December 13, 2011


Speaking of deceptive graphs, I liked this.

That's not deceptive so much as just plain plotted incorrectly/lying.
posted by jimmythefish at 7:06 AM on December 13, 2011 [2 favorites]



"It's all about juking the stats"
posted by Ruthless Bunny at 7:07 AM on December 13, 2011 [3 favorites]


Well, to be fair I know at least a few people who were all "WOOOHOOO we bought a house, we never thought we'd pay so much! Also, we're having a baby and we will name her Ava!"

I'm not joking.
posted by jimmythefish at 7:13 AM on December 13, 2011 [2 favorites]


Speaking of deceptive graphs, I liked this.

That's an incorrect graph. The OP's graphs are (presumably) plotted correctly.

The best defense against the OP's graphs is to read the paper first -- they must prove that there is some causal connection between, say, dollars spent on Milky Way candy bars and the supernova rate in the Virgo Supercluster before they can use a chart to help demonstrate that the cause occurs in line with the effect.

Or, if you wish, you can show a chart stating X->Y once you show how X->Y. Then a chart becomes very useful in showing the extent of X, and thus, Y.

I can find lots of coincidental shapes. It's even easier if I don't have to have line graphs X axis at 0, have all plots on the axis use the same origin, use linear scales for some lines and log scales for others, use unrelated units, and so forth.
posted by eriko at 7:38 AM on December 13, 2011


Oh I'm dreading the moment when that climate change chart is repeated free of context.
posted by Navelgazer at 7:43 AM on December 13, 2011


Speaking of deceptive graphs, I liked this.

That... I... what? I don't even
posted by EndsOfInvention at 8:01 AM on December 13, 2011 [1 favorite]


I'm just glad that finally, someone online recognized that correlation does not equal causation.
posted by a snickering nuthatch at 10:13 AM on December 13, 2011 [1 favorite]


"I used to think correlation implied causation, then I took a statistics class and now I don't."

"Sounds like the statistics class helped."

"Well, maybe."

xkcd
posted by jasper411 at 10:30 AM on December 13, 2011 [2 favorites]


That's fun and interesting and all, but honestly.......what the heck happened to the murder rate in NYC?!
posted by AbnerDoon at 11:22 AM on December 13, 2011


A mountain range. Didn't you read the link?
posted by arcticwoman at 11:31 AM on December 13, 2011 [5 favorites]


My age; inflation.

DO THE MATHS, PEOPLE.


So... clvrmnky, your age fluctuates, and you actually got younger in 2009?
posted by IAmBroom at 1:01 PM on December 13, 2011


Some random chartline, scaled to match the reality of the geographic landscape of...where, exactly?

I'm guessing that the mountain range is photoshopped to match the data, like what Michael Najjar has done with stock market data (examples).
posted by -jf- at 5:21 PM on December 13, 2011


LOL
posted by Saddo at 3:18 AM on December 14, 2011 [1 favorite]


Don't these graphs take it one step further? I don't think they even show correlation, which requires some inter-variable dependence.
posted by benbenson at 9:43 AM on December 14, 2011


Well, they're both in the same chart together. I think that makes a solid argument for their dependence on each other to demonstrate the causation.
posted by iamkimiam at 5:37 PM on December 14, 2011


« Older Come and see for yourself ...   |   Wherefore art thou (probably), Higgs? Newer »


This thread has been archived and is closed to new comments