Zachary Goldfarb examines the racial disparities in the JAMA study:
We’re celebrating the fact that for all kids ages 2 to 5 childhood obesity has declined from 13.9 percent to 8.4 percent over 10 years. Yet, 11.3 percent of black children ages 2 to 5 and 16.7 percent of Hispanic children that age are obese. Just 3.5 percent of white children ages 2 to 5 are obese.
Why the disparity? Income certainly plays a central role, though this study didn’t look at that factor. Researchers have other ideas, including the fact that black and Hispanic children eat solid foods earlier than doctors recommend, watch more television, have a higher intake of sugar-sweetened and fast foods and have mothers who face higher levels of maternal depression. It’s not a hopeless situation – breastfeeding by black and Hispanic children has increased, and government programs are fighting the disparity — but the gaps are vast.
Meanwhile, Razib Khan casts doubt on the study. A reader joins him:
In your post about the JAMA study about kids slimming down, the caveats loom large.
In a nutshell, in the minds of the authors themselves, the results of this study are sufficiently shaky that they are reporting only a modest decrease. According to the authors’ own statistics, the data is sufficiently noisy that such a decrease isn’t particularly surprising [1] even if the observation were the only one they examined. However, there is an even more disturbing shortcoming of the study’s conclusion about kids, one which the authors again acknowledge [2]. While it may be at most mildly surprising to observe that large of a change if you only made one observation, if you made a lot of observations, you’d be terribly surprised if a handful weren’t that extreme by chance alone. Surprisingly, JAMA allowed this overhyped conclusion based on an unsurprising and marginal result to be published. Predictably, it has generated headlines around the web. While predictable, this is very frustrating to scientists like me.
[1] In their paper, they write:
There was a significant decrease in obesity among 2- to 5-year-old children (from 13.9% to 8.4%; P = .03) […]
In statistical parlance, they observed a p-value of 0.03 or 3%. This means that, even if there were no true change in rates obesity, noise alone could underly an apparent decrease of this magnitude 3% of the time. In biomedical studies, the most speculative and lax standard for calling a result “statistically significant” is conventionally set at 5%. There is nothing magical about 5%, but it comports with our sense of “not very likely”. Suffice it to say that most scientists consider 5% significance to hardly worth mentioning. See here, here, and here for how physicists deal with the issue of “statistical significance”.
[2] The authors admit:
In the current analysis, trend tests were conducted on different age groups. When multiple statistical tests are undertaken, by chance some tests will be statistically significant (eg, 5% of the time using α of .05). In some cases, adjustments are made to account for these multiple comparisons, and a P value lower than .05 is used to determine statistical significance. In the current analysis, adjustments were not made for multiple comparisons, but the P value is presented.
Why this didn’t undermine the whole argument in the minds of the reviewers and/or editors of JAMA is beyond me. As a reviewer, I certainly would have been been extremely skeptical of one or a handful of marginally statistically significant results, especially when the 3% result is the one generating the biggest headlines. That 3% is virtually indistinguishable from a 6% result result that wouldn’t have generated headlines.
But nevertheless, Paul Campos feels that the JAMA findings are “consistent with broader international trends”:
As Michael Gard notes in his recent book The End of the Obesity Epidemic, data from all over the world indicate that, over the past ten to 15 years, obesity rates have leveled off or declined among adults and children. This is an awkward development for obesity fear-mongers, who as Gard and others have pointed out, have repeatedly claimed that within another generation or two the entire population of some nations, most notably the United States, would be fat.
The claim that obesity is an epidemic phenomenon, and that its prevalence was on the way to approaching 100 percent, has always been crucial to the other claims of the anti-fat industry, most notably that today’s children will have shorter lifespans than their parents, and that obesity is as great a threat to public health as global warming. There has never been any real evidence for these claims: life expectancy continues to rise, overall health continues to improve, and today’s young people are healthier than their parents were at the same age. Faced with these inconvenient facts, the obesity police have always argued that, while it’s true we’re not seeing the supposedly devastating effects of a heavier population at the moment, we will surely see these effects if present trends continue.
The Dish 