Maybe Freakonomics has made me attuned to issues of statistical aggregation and receptive to counterintuitive claims, but two items recently have leapt out at me.
First, Carpundit makes the claim that increased speed limits do not lead to increased loss of life. It’s an incredible claim, though not necessarily wrong for it. I, like many, was under the impression that studies had found correlation between speed limits and accident rates. Am I wrong? Or were there reasons said studies don’t show a true causation?
Second, Monica Collins spends an entire essay pondering a Cambridge Health Alliance study showing a statistical correlation between women who smoke or drink heavily and those who experience domestic violence. Now the logical response would be that a correlation does not show the cause of domestic violence, that both smoking/drinking and domestic violence owe their source to some other third variables. (Collins says the obvious culprit is “inner turmoil”, whereas I can think of a host of variables that would be more empirically measureable.) So Collins hedges at the end, unsatisfactorily:
The new study does not say cigarettes and alcohol lead to domestic violence. But it sure is sobering, making all that cultural imagery of taking a swill and striking a match and a pose look even more dated and deceptive. Romance vanishes when Bette Davis’s eyes turn black and blue.
Huh? If the study does not prove or even indicate causation, then why is it sobering? I mean, domestic violence itself is sobering, but it’s no more prevalent than we thought before the study. And we have no indication that imitating the poisons of choice of onscreen characters is going to lead anyone to get beaten up.
No comments have been added to this post yet.