Comment by yolo69420 | Hacker News Reader

Comment by yolo69420 | Hacker News Reader

Parent Full thread

I don't quite think you understand. The data presented here absolutely does not allow for the conclusions you make particularly in your second paragraph. In fact the reality could be exactly the opposite way (i.e. geography has no influence on things like modern election results whatsoever) and you wouldn't know it based on this data.

Think about the countless numbers of ways you could overlay any sort of historical map over some map representing any kind of relevant statistic. A smaller but still countless number of them will have strong correlations by pure chance alone. This is a simple fact of statistics.

The author here is filtering out exactly those that happen to have those kinds of correlations backed by some preconceived ideas the author has about how the causality is supposed to work and uses them to fuel his hypotheses, while ignoring the uncountable number of map overlays that don't show these correlations.

This is bog standard selection bias.

You're correct. I was saying that correlation implied either causation or at least a common cause, and you correctly point out that there can be (edit: and frequently are) spurious correlations. The desire to see causal effects is strong in us humans, and it is easy for us, me especially, to fall prey to this logical fallacy.

For people who would like to see a demonstration of the effect yolo69420 is pointing out, you can visit the website below [1]. It, among other things, shows how Nicholas Cage movies are correlated with swimming pool drownings.

I completely agree that the Twitter post does not prove there is a direct causal relationship between the maps they compare, nor even that there exists a common cause. I would contend that this is a very high standard to hold for interesting content on the internet. If you would like to hold to this standard, you will likely come to more robust and defensible conclusions than myself.

I find it fun to speculate on possible common causes. If I had any actual ability to change things based on my conclusions, I'd likely increase my burden of proof to a much higher level like you suggest.

[1] http://www.tylervigen.com/spurious-correlations

Reply on news.ycombinator.com