Causation or correlation?

So here I am heading home from the office the other day listening to All Things Considered on the radio and there's this story about Kenya trying to fight the scourge of HIV by encouraging men to get circumcised. The reporter quoted a scientific study that the rate of HIV infection in uncircumcised men was substantially higher that the rate for men who were circumcised.

Well, that got me to thinking. This link between circumcision and HIV infection rates - is it an example of causation or correlation?

In other words, does getting circumcised reduce one's risk of getting infected by HIV, or is there something else at work? Is it because uncircumcised men have riskier sex practices? Is it because circumcised men are practicing safer sex?

The literature appears to be all over the place. Of course since the World Health Organization (WHO) jumped on board the circumcision train money has poured into Africa to fund circumcision programs. But there is still precious little hard evidence that circumcision has anything to do with HIV prevention.

Adding to my skepticism that there is no causal relation between the two are activists in Europe and the United States attempting to ban the procedure on newborns. If circumcision really prevented HIV infections, why would anyone be campaigning against it?

Now I'm no doctor and I am not going to sit here and pretend that I have some vast store of knowledge about the transmission of infectious disease or the human reproductive system. But I do think this issue is a perfect example of the causation v. correlation puzzle we run into in the courtroom.

The police coordination exercises the police use on the roadside to determine whether they're going to arrest a motorist for driving while intoxicated have everything to do with correlation, not causation. The studies used to provide a modicum of "scientific heft" to the roadside exercises all speak of correlations between levels of intoxication and performance on the exercises. Not one of them show evidence of causation.

The distinction is important. Let's say you're watching your favorite team play in a crucial game (sport and team not important). All season long whenever you've worn a certain shirt or cap or drank a certain beer or ate a certain sandwich, your team came through. Your actions did not cause your team to prevail. There was no causal link between your shirt (or your cap or beer or sandwich) and your team's performance. There was just a correlation - when you wore your shirt, your team just happened to win.

On the other hand, if it's dark out when you're driving home you turn on your headlights - as does everyone else out on the road. The lack of light caused you to turn on your headlights. There is a causal connection between nighttime driving and headlight use.

We can draw correlations between all sorts of statistics but just because you can draw a correlation between two observations doesn't mean that one caused the other. Your lucky socks won't help you win that poker game tonight, but your skill in reading the other players and their cards will.

Just because a motorist exhibits certain clues or signs or behaviors when walking up and down a straight line, doesn't mean that alcohol caused his performance. There are plenty of other factors that are not taken into account by the officer asking you to perform the exercise. And while the so-called validation studies may have drawn correlations to intoxication and performance on the roadside exercises, the studies did not make any causal connections.

The prosecutor will try to make it look like there is a causal connection between intoxication and performance, it's your job to point out to the jury the difference between causation and correlation.