Friday, October 4, 2013

Pit bull advocacy and basic statistics

There have been 400 disfiguring or fatal US pit bull attacks on humans so far in 2013.

In the aftermath of each attack, news sites that report the attack are invariably besieged by an army of relentless pit bull advocates doing damage control and PR, delivering prepared and well-practiced talking points which usually boil down to one or more of the following, in some form:

  • It was the attack victim's fault; It was the pit bull owner's fault
  • Blame the deed, not the breed! (What does that even mean?)
  • How do we know it was a pit bull? (It was a pit bull, not a unicorn!)

One of the most popular types of talking points involves a personal anecdote, usually involving one or more of the following elements:
"I've had pit bulls all my life and I've never been bitten by one, but I was viciously attacked by a <arbitrary breed name here>. My pibble is harmless. He wouldn't hurt a fly."

The pit bull lobby always tries to emphasize atypical behaviors - for instance, a seemingly mild-mannered pit bull. But when a pit bull advocate trots out an anecdotal tale of a pit bull that didn't attack, thinking it to be some sort of game changing revelation, what they're really saying is that they have no concept of statistical distribution.

While there may well be some "family pit bulls" that don't attack, there is absolutely no way to predict whether such a pit bull will attack tomorrow, since virtually all attacks by such pit bulls are completely random, sudden and unexpected. But let's say, for the sake of argument, that there are pit bulls that don't attack. This is no surprise to anyone who has been paying attention. Living creatures are not identical automatons. For every trait, there is variation along some curve. Pointers were bred to point, but you might occasionally find a pointer who fails to point. Pit bulls were bred to attack, torture and kill weaker creatures for no particular reason, but you might find the odd pit bull that doesn't appear to have that inclination. 

At any rate, the pointer that doesn't point, the lab that doesn't swim, and the pit bull that doesn't torture and kill are statistical outliers. The fact that such pit bulls might be observed is often used by pit bull advocates to insinuate that all pit bulls are harmless, but all it really does is illustrate the properties of the well known bell curve.

The bottom line is this: Any particular single observation or measurement could be nearly anywhere on the map. A single data point is always inconclusive. But once you have a large enough sample size, every distribution starts looking like a bell curve. (More info on this at Khan Academy)

So, when pit advocates trot out the apocryphal account of a meek pit bull (assuming it's true in the first place) we're talking about something not representative of the breed in general. When they trot out their tale of an attack by an aggressive large retriever (again, assuming the story has any basis in fact) it is a statistical outlier. When pondering what breed of dog you would like to adopt as a family pet, it would be extremely foolish to focus on an aggressive breed and assume you'll end up with one of the safe "duds", and it would also be a mistake to avoid a known safe breed because you heard a story about someone being bitten by one. 

As long as we're dealing in personal statements about pit bulls, consider this observation from a long time pit bull fancier and breeder: "Any pit bull that hasn't killed another dog is a pit bull that hasn't been let outside". Isn't that lovely? The pit apologists will howl in protest over that statement, but it has every bit as much validity as the hackneyed "my pit bull wouldn't hurt a fly" shtick. 

Anyone can make a statement of belief, and anyone can relate a personal anecdote, but if we want an accurate view of reality, we need to look at the big picture, not just our own personal experience. Using a search engine can be a useful first step in such a quest. Google is your friend!