Skip to content

Comment on Storks Deliver Babies (p= 0.008) (2000) [pdf]parent

Comments

Also related: Spurious Correlations[0]. These are great and really drives home the idea that correlation neither equals or implies causation. At least one of these graphs implies that if the US wanted to reduce the number of drivers getting hit by trains, they simply just need to stop importing oil from Norway.

[0]: https://www.tylervigen.com/spurious-correlations

There are some cases where lack of correlation is actually evidence of causation.

As in the causal effect happens to cancel out a confounder, or something more fun? Do you have an example?

I was thinking of an air conditioner - a perfect A/C would mean the temperature is flat. Although hmm, I think R^2 is undefined rather than 0 there.

But basically, it's a "if you do it right, it'll look like you've done nothing at all" situation.

What variables are you correlating in that example? It seems to correlate just fine if your measuring energy inputed to heat the room vs work done by air conditioner.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.