🚀 go-pugleaf

In sci.skeptic troppo <troppo19@gmail.com> wrote:
...
> Also, other articles seem a bit easy-going in attributing cause and effect. 
> Related, yes. Causally related? Very difficult. Coefficient of 
> Determination can, and has been, 98.5, and there it turns out that there is 
> no causal relationship. The old problem of 'ice cream causes drowning'.
...

[For the original s.s audience, perhaps:]

For pedagogic reasons maybe better explained as the "hidden variable" 
problem. I.e. if 2 things appear significantly related we can't be sure 
without any a priori idea of causation they are directly connected, or 
whether there is a 3rd factor (the "hidden variable") that causes both of them.

I.e. in the case you mention "hot weather" would a priori explain an 
increase in ice cream consumption and an increase in swimming and 
(therefore) drowing. Hence the 2 would appear related. And they are. :)

Direct causation can never be easily ruled out, anyway.  Eating icecream 
before swimming may induce cramps or other physical effects and therefore 
influence drowings, too.  A careful analysis with enough data should 
disentangle the various links.

Of course it's better to start with assumptions/a model of causal
links and THEN get the conf intervals and r2's to maybe decide whether it's 
relevant, and how much so in comparison to other relevant factors.

-- 
[confusing the years between records and the number of years the 
records are about:]
And BTW, by your logic, the 1998 hot year (supposedly hottest EVAHH) 
just proved hot years are rare. Does that sound right to you?
Yes, take time and think it through.
  -- Ron House <rhouse@smartchat.net.au>, Bris, 09 Dec 2010 23:31:38