PDA

View Full Version : Astrological "evidence" produced by looking at the tails of a normal distribution


Jack21222
Feb5-12, 09:42 AM
I just ran across this on a large news aggregation website, so I'm sure some people here might have seen it as well.

http://www.inquisitr.com/173752/jail-time-predetermined-by-astrology-signs-says-police-study/

But just because the police don’t see the value in such a list doesn’t mean that the pattern isn’t there. Georgia Nicols, who writes the National Post’s horoscope, said that it isn’t surprising to her that Aries are arrested the most.

“Aries is the sign of the warrior. Aries rules the military. Aries jump in head first, and love adventure. A lot of people in the newsroom are Aries.”

If you look at the data, it appears to be a gaussian distribution, which shouldn't be too surprising. They just picked one of the tails (and it isn't even an extreme outlier, maybe 1 or 2 sigma, I haven't done the math) and are trying to attach some significance to it.

This is commonly known as the Texas Sharpshooter Fallacy (http://en.wikipedia.org/wiki/Texas_sharpshooter_fallacy), where this article is drawing a bullseye around the outlier.

I'm sure this isn't the first time people have tried using outliers in a normal distribution to show evidence for astrology (or any number of other woo-woo claims). So, it's important to keep your tools sharp in your baloney detection kit, and keep an eye out for this sort of thing.

Vanadium 50
Feb5-12, 05:47 PM
The probability that the worst month will have 203 or more arrests is calculable, and is 3.4%. I can't get excited about that.

Moonbear
Feb5-12, 08:38 PM
I'm willing to bet that if they tracked data for multiple years, the annual variation would be similar to or exceed variation based on birth date.