The Obesity Rate for Children Has Not Plummeted

People misunderstand this stuff a lot. Frequently, it leads to them overestimating risk.

For example, my wife keeps up with a million different organic food and cancer prevention-type blogs and news feeds. She'll read something like, "People who eat cilantro in their rice were shown in one study to have a 3 percent greater cancer risk."

Of course, it'll be one study in 1,000 on that topic and the only one that showed this connection. And the 3 percent greater risk will increase the risk from 6 people per 1,000,000 to like 8 people per 1,000,000,000. But from then on, cilantro will disappear from the shelves at Whole Foods - and our refrigerator.

I am joking about cilantro, but I think if you Google Vitamin D and cancer, you'll see a bunch of freak-outs about some outlier study.

 

I'm doing a piece right now on effect size and publication bias (the idea that scientists publish "significant" results and file-drawer non-significant ones). Especially in the social sciences, effect sizes can be very, very small. What I've read re: organic/natural food is that the same is likely true.

 

That'd depend on how you measured the "development of a lifetime risk," I'd guess, along with mortality rates for various types of it, which I don't know.

 

If he didn't argue the same thing, it's probably true that a reasonable inference could be drawn from his work to support such an argument. Nevertheless, I am not arguing that that happens in my piece. Rather, I am accepting it as a given and pointing to approaches to quantify -- perhaps underscore might be a better word -- just how small effect sizes can be.

 

doc, with all this numbers knowledge you need to do something useful with your life like get over to the strikeouts thread.

 

The government in charge will spin results any way they want to support its policies.

 

Well, there's also the effect on the career of him/her who's doing the publishing.