Thursday, April 23, 2009

Correlation does not mean causation

This is the mantra of statistics professors everywhere. Here's an example: City populations with 100' wood telephone poles have a higher rate of heart attacks, therefore 100' wood telephone poles cause heart attacks. This is a causality fallacy - I don't have any evidence that height or material or telephone poles actually CAUSE heart attacks, I just have a correlation. Events that are correlated can certainly be caused by one another. But this is only fairly established (except for the most obvious things, like how stepping in front of a 18-wheeler will cause you to go splat) when confounded variables are eliminated and more research is performed. For a nice discussion on correlation and causation, check out George Mason University's

What does this have to do with dogs or pit bulls? I've found that the media is very good at engaging in causality fallacies. (I'm not innocent of falling prey to this logical fallacy, it's hard not to sometimes). For example, when media reports on the results of studies that show a high correlation of something, often reporters claim a causation when there isn't any evidence to suggest that. It's a lot easier to say "x causes y" than "x is correlated to y and, without a preponderance of more evidence, we cannot say that x causes y". Reporters don't always do this, as there are plenty of studies that establish very strong correlations between two things, enough to suggest a causality in lieu of pure coincidence or another relationship.

As to dogs, here are a couple of recent examples of published causality fallacies.

Code Compliance Director Brandon Bennett told Fort Worth city council officials that the growing number of aggressive dogs on the street didn't translate into all pit bulls are aggressive (there also was no data on what "growing number" meant).

In response to this statement Star-Telegram reporter writes "But the number of pit bulls and pit-bull mixes brought into the shelter grew to 4,631 in 2008 from 2,154 in 2004."

The Star-Telegram seems to be suggesting a lot of things with that statement, including that the increased intake of pit bulls at a shelter is a) causing an increase in aggressive dogs and b) indicating, well no Code Director Bennett, all pit bulls are aggressive - see how they are flooding the shelter with their aggressiveness? There isn't any evidence that increased intake of pit bulls causes anything but an increased intake of pit bulls.

Another causality fallacy comes from the IndyStar reporting on the decision made to start adopting out pit bulls from the shelter.

"Critics feared the policy change would increase attacks, but so far that does not seem to have happened.Animal Care and Control has responded to 34 pit bull bites since the shelter started adopting out the breed. During the same period last year, the agency responded to 46 pit bull bites."

I like the sentiment but it's off-base. There is no evidence presented that the act of adopting a dog from a shelter causes an increase in dog bites. I don't have data to support this, but I imagine shelters would be hard-pressed to continue adoption programs if there was a causal relationship between adoption and dog aggression towards humans (lawsuits and medical costs alone would be enough of a deterrent).

So while it's nice that a case is made that pit bull bites have decreased, there is no logic in claiming that had bites from pit bulls increased, that increase would have been caused by Indy Animal Care and Control adopting some pit bulls out (excepting in the very unlikely event that ALL the pit bulls adopted out were responsible for the increased bites).

I'll try my best not to fall prey to this logical fallacy, though it seems like reporters and government officials might want to work even harder. :)

No comments: