# "Pure Lunacy"

Written by Paul Sutter on Monday, 03 July 2017. Posted in From The Desk of...The Chief Scientist

Let's imagine some report comes out claiming that incidents of crime are higher than average during the days surrounding a full moon. And let's further imagine that there are enough data to claim statistical significance. What are our options to interpret this result?

a) The full moon causes people to go nuts because of mysterious moon-rays b) The full moon is connected to people going nuts for some mundane reason c) None of the above

I suppose it's easy to dismiss option a) since we don't encounter crazy-making moon-rays in any other experiment, so most rational-thinking people rightly go to conclusion b). The moon itself isn't causing people to go berserk, but perhaps with the slightly brighter nights, folks are more inclined to think criminally.

Perhaps. But to me the simplest answer is c). If you mine enough data and make enough cross-correlations, you're bound to find a statistical link - even a significant one. But "statistically significant" is not the same thing as "meaningful". You can see this in action on one of my favorite websites, Spurious Correlations: tylervigen.com/spurious-correlations, which finds significant connections between, for example, the number of swimming pool drownings and the number of films starring Nicholas Cage.

Be careful with statistics - data are the first things to lie to you.