Hmmm...

via.

Comments

[Dan Richter]

Man müsste noch eine Statistik haben von rauchenden und nicht rauchenden Dicken.

[Norbert]

Irrespective of the extent of causality, perhaps the conclusion to be drawn from this graph is: since society successfully managed to make smokers outcasts in less than 30 years, it should apply the same strategy to obese people? After all they do place a burden on society as a whole....

[Andrew]

Actually, the most accurate warning is 'Correlation is not necessarily evidence of causality.' But it certainly can be.

[christian]

What Andrew wanted to say with "Hmmmm" was actually:

"Fat fingers make it harder to pull cigarettes out of their little boxes which is why obesity leads to less smoking. Given that correlation proves causality, this graph settles the matter once an for all."

[orangeshow]

The Euro's always say "American's are fat." The American's always say "All the Euro's smoke, it's disgusting." The R-squared coefficient you have just displayed here = 0.99 !!! (Statistically speaking, of course).

[G]

Perhaps you should look for the addition of HFCS to everything (including bread, biscuits, soda) to see some actual causality.

[LemmusLemmus]

People, just chanting "correlation is not causality" does not betray an in-depth understanding of the matter. While one is well-advised to look for "hidden parameters" that drive a correlation, I would like to see spelled out the argument that the passing of time, in and by itself, drove the decrease in smoking or the increase in obesity.

Of course one can make all kinds of nonsense graphs representing noncausal correlations, but the difference here is that it is widely accepted that smoking causes weight to be lower than it otherwise would be (as experienced by many people who quit smoking). As one review paper* put it:

“[S]ome adolescents may be influenced by their beliefs that smoking can control body weight. [...] In fact, smoking does suppress body weight (Williamson et al. 1991), which makes this attitude particularly difficult to counter.”

It would thus be highly surprising if the decrease in smoking didn't play a role in the increase in obesity.

Not that I would translate "Hmmm..." as "Given that correlation proves causality, this graph settles the matter once an for all" in the first place.

_____
*Timothy B. Baker, Thomas H. Brandon and Laurie Chasson, 2004: “Motivational Influences on Cigarette Smoking”, Annual Review of Psychology 55: 463-91. The quote is from p. 469.

[christian]

Turn any well-established rule, e.g. that pregnancy is caused by sex, into a national statistics and people will turn up who insist that it is actually sex that is caused by pregnancy or that sex and pregnancy are both caused by a third, yet unknown factor.

[M.R.]

my all-time favourite:
http://www.scienceblogs.de/primaklima/carteFN.gif
(Tschernobyl -> voting for Front National)

[N]

You could make the same graph with the US debt, giving you the equation more debt = more fat.

Or the number of cars in asia, for what it's worth.

Apart from that: correlation is not causality.

[surfguard]

Repeat after me: Correlation is not causality. Correlation is not causality. Correlation is not causality. Correlation is not causality. Correlation is not causality.

(Hint: Look for the hidden parameters that drive both factors and keep in mind that time is always one of them.)