Wednesday, August 06, 2003

Collapsing Buildings, Babies, and BS: Okay, Razor, there may be something to the study -- it's in a reputable, peer-reviewed journal, after all -- but it looks pretty suspicious. Not the study itself, but the media reaction. From the Newsday article you linked:
A preliminary report on 187 pregnant women who were in or near the collapsing towers shows that the soot, pulverized glass and other toxins proved so detrimental that mothers who inhaled the debris delivered babies who averaged a half-pound lighter than infants of unexposed women. Researchers studied women who were within a half mile of Ground Zero on the day of the attack or during the succeeding three weeks.
The report "shows" no such thing. Even without reading the report, I'm certain of this. Why? A "preliminary report" could really only reliably show correlation, not causation; with a sample (n) as small as 187, you're getting into a realm where typical statistical tests are not reliable. You see this in political polls all the time: the smaller n is, the higher the margin of error. (And any political poll in which n=187 [as a sample of an indefinite, but certainly huge, population] would be given zero credence by anyone with a stake.) Furthermore, what confounds are left unexplored? How do you factor out variables like stress, which can have a powerful effect on the biochemics of the body? I guarantee that the authors of the report itself use tremendously more circumspect language, specifically avoiding words like "shows" or proves" when discussing the data. Newsday just wants a splash headline.

More: David Bernstein, blogging at the Conspiracy, rewrites the NYT headline for this story. To be fair, once you get past the headline, the NYT story is much more cautious:

New York City health officials said the implications of this kind of study were limited by its size and methods, which, among other things involved relying on the recollections of the women being monitored.
And, in answer to my question, they did try to control for stress factors:
Any psychological trauma from directly witnessing the attack or its aftermath appeared to be ruled out as a potential cause because the two groups were given surveys that showed no difference in their tendency toward post-traumatic stress disorder, said Dr. Philip J. Landrigan, an author of the study and chairman of the department of community and preventive medicine at the school.
I guess the NYT is still good for some things.

No comments: