Tag Archives: selection bias

Seems, madam?

Weaving nutrition and exercise lessons into middle-school classrooms can reduce eating disorders among girls and ultimately save medical costs, a study by Boston researchers concludes.

The researchers analyzed data from an earlier study at 10 Massachusetts middle schools, including five that adopted an obesity prevention program called Planet Health, and five that did not.

…

Austin’s team estimated that if the obesity-prevention program they studied was expanded to 100 schools, there would be about $680,000 in health care savings.

From the article “Nutrition lessons for girls seem to pay off” in the August 2 Boston Globe. Commenters note small sample size and that this seems like an ad for “Planet Health” instead of an unbiased article.

Are women more likely to “sext” than men?

2 Replies

Yes, says this piece in the Fashion Section of the July 17 New York Times.

At least according to a recent study by Diane Kholos Wysocki, a professor of sociology and women’s studies at the University of Nebraska at Kearney, and Cheryl D. Childers, a professor of sociology at Washburn University, women are more likely to send nude photographs or sexually explicit text messages than men. About two-thirds of women surveyed sent such missives compared with roughly half the men (although for reasons to be explained later, these results are not the most scientific).

Why aren’t these results “the most scientific?”

But there are major caveats to the study: Surveying users of a Web site for swingers does not compare to taking a nationally representative poll or conducting a community study. The people who responded to the survey differ from the general population in many respects; most significantly, they are actual or would-be cheaters. They are also people who visit a particular cheating Web site, which may differ in clientele from other cheating sites.

Finally, the respondents were people who elected to respond to a voluntary survey and therefore differ from those people who visit that Web site but do not have the time or the inclination to fill out a survey. This is called self-selection and makes generalizing about the entire population unreliable. In the end, the group skewed more educated, more affluent and older than the general population.

Still, the study does offer insight into the world of online infidelity.

I know the New York Times can never print an article that doesn’t include a paragraph starting with “Still,” but come on. What insight do we get into anything, with all these caveats?

I also like the bold heading at the top of the page: “THE GIST Women are more likely to “sext” than men.”

We report, you decide if you should drink and drive?

RESEARCHERS AT the University of California, San Diego, use questionable data to make the flimsy argument that a few sips of alcohol are enough to significantly increase the likelihood of serious injury in an automobile accident (“Study: 1 drink raises driving risks,’’ g section, June 27).

Their data only reflect numbers for accidents in which there is a fatality. That’s only three-tenths of 1 percent of all car accidents – 34,172 out of 10.2 million. It’s like trying to discover how widespread steroid usage is and using Major League Baseball players as your sample.

I’m too tired to evaluate this. Any takers?

Returns to college education

David Leonhardt of the New York Times jumps in to the emergent public policy debate over the value of a college education. The short version of that debate: college is expensive, and it’s not clear everyone benefits from it, so maybe we should stop encouraging everyone to go.

I probably agree with Leonhardt, but think he could make a better case. His main piece of evidence is that even people in “lower-skilled” occupations make more with a college degree than those who don’t. Witness his headline (which, to be fair, he probably did not write): “Even for Cashiers, College Pays Off.”

He cites a number of studies, but unfortunately they all have big problems, problems that Leonhardt even seems to realize himself:

When confronted with such data, skeptics sometimes reply that colleges are mostly a way station for smart people. But that’s not right either. Various natural experiments — like teenagers’ proximity to a campus, which affects whether they enroll — have shown that people do acquire skills in college.

Well, that’s not really right either. These “natural” experiments have their own problems, and it would be nice to talk about those too, along with the substantive magnitude of the effects. Like, when we correct for the “way station” argument, how much does the effect go down?