Cigarettes, damn cigarettes and statistics | Hacker News Reader

Cigarettes, damn cigarettes and statistics

timharford.com

30 pointsPetiver14 comments

"Correlation is not causation", while true, is becoming a very tired response to pretty much any suggested phenomenon here on HN. And it shuts down often worthwhile "what if...?" conversations.

This is especially well put:

"We cannot rely on correlation alone, then. But insisting on absolute proof of causation is too exacting a standard (arguably, an impossible one). Between those two extremes, where does the right balance lie between trusting correlations and looking for evidence of causation?"

That's what I was going to point out

"We should respect correlation but it is a clue to a deeper truth, not the end of our investigations."

It is a clue. It might mean nothing but it might lead onto something.

But of course a lot of people think they're smarter by quoting a phrase.

"Anecdote is not data" but it was anecdotes about people getting cowpox not getting smallpox that led to the development of immunization.

Exactly. Perhaps it's time for a new catchphrase:

"Correlation not being causation does not require dismissal"

But it might just lead to:

"Correlation not being causation not requiring dismissal does not imply truthiness"

and we're back where we started :-)

"Correlation is not causation" is still a valid comment to make, the original article says that deeper questions need to be asked.

However most of the time a news report just reports a single correlation. That's when those comments come out, its statistically insignificant to have one source/correlation and report it as news. Show me two correlations or more and I'll take more notice.

For example:

1, Smoking causes cancer 2, Smokers have a higher rate of lung cancer 3, Heavy smokers are more likely to develop lung cancer sooner than casual smokers 4, Smokers that quit lower their chances of cancer

As a set they make an valid argument, however individually they are weak.

I'm getting tired of seeing "correlation is not causation" thrown around all the time, too. It's so easy and smart-sounding, saying it is guilt-free and feels good too.

It probably is too high a standard. But perhaps it is also an indication that our ability in proving causality needs to be enhanced. It's incredibly difficult---can it be made easier?

Sidenote about tobacco:

Each time I walk out in the city, and somebody is smoking around me and the smoke reaches me, I feel like I need to hold my breath and change course to absolutely avoid breathing that smoke. Feels like poison.

I guess it's psychological, but smoke seems to make me feel sick, even at low doses. Even the smell of smoke from the neighbor's window will upset me.

I guess I either am a victim of anti tobacco ads, or that smokes is a real legal poison that my body can't sustain.

Yet the traffic fumes won't bother you. I have seen a few places where they say the pollution is equivalent to almost a pack of cigarettes.

http://news.bbc.co.uk/2/hi/uk_news/scotland/4401628.stm

I feel the same as GP, and it extends to pretty much all forms of smoke. Car (or worse: scooter) in front of me while cycling often makes me choke and need to slow down to catch my breath again. Incense works as well. In my case it's very likely a psychological hypersensitivity, although given that smoke in any form is rarely healthy it's not something I'd go to great lengths to fix.

Barcelona usually has a fair bit of wind, but when it stops blowing the traffic fumes are terrible (as a cyclist). And I noticed that as a smoker.

Yes they do bother me. Diesel especially. Scooters too (two stroke engines pollute more).

Interesting that Judea Pearl's work on inferring causality from data is not mentioned. http://bayes.cs.ucla.edu/BOOK-99/book-toc.html

And don't forget the later "Causality" (2nd edition): http://www.amazon.com/Causality-Reasoning-Inference-Judea-Pe...

I like how every time someone tries to defend the status quo of swallowing correlations pretty damn naively, they use smoking as an example; apparently the best example dates from ~70 years ago, and they completely ignore the continuous din of medical failures where correlative results failed to transfer to randomized trials (most prominently, anything to do with diet).

Reply on news.ycombinator.com