Discussion about this post

User's avatar
anon's avatar

> a change in the rate of brain in obesity rate is expected

lol

Expand full comment
Alex DeLarge's avatar

I think it is slightly confusing as to what this statistical phenomenon means in practice. If I understand, there are two interrelated points being made:

1. Because a normal distribution is thin at the tails and thick in the middle, it follows that the % of the curve which is to the right of a right-of-center threshold increases more than the mean itself when the mean shifts to the right.

Ok. So this means the percentage of arbitrarily defined "overweight" people (BMI > 30) increases faster than the average BMI of the average population. This is definitely an interesting observation, but it doesn't make the "increased obesity %" stat wrong in any way. The "BMI > 30" stat never purported to measure the population's average body fat or whatever.

You say the “BMI is a quite decent predictor, or proxy of, body fat %.” And it looks like BMI 30 = around 30% body fat for men. So the issue here should be whether this BMI 30+/30%+ body fat threshold is meaningful for health (or even aesthetic) outcomes? In other words, how does, say, all cause mortality scale with BMI? Maybe 30% body fat is an important inflection point. Or maybe it isn't and the health effects are just linear.

In other words, if the increase in bad health health effects scales at a non-linear rate that is the same or greater than the non-linear scaling of the >30% body fat metric, then it may be entirely appropriate and most accurate to define the "obesity epidemic" by this "arbitrary" metric.

2. If mean BMI has necessarily risen more slowly than the BMI > 30 threshold that has implications for dating the overall increase in BMI and looking for possible environmental causes.

Ok. That seems like an important point. But does that mean that the accelerated increase in the rate of individuals with BMI > 30 is unimportant?

I don't know much about physiology, but I could imagine that there might be some homeostasis effects that might self-regulate hunger and body fat so that population BMI should be "anchored" around some objective healthy level rather than just varying around an (unhealthy) population mean. Even from just 2000 to 2018 you note that the BMI curve is slouching to the right, indicating that the ratio of extreme fatties is accelerating, with a big increase in SD "from 6.6 to 7.8." (Which is seemingly being generated solely by the increase in the right tail of the BMI curve).

In other words, isn't that change in SD prima facie evidence of some real phenomenon that is disproportionately hitting the right tail of the curve? (Or is this a statistical artifact of how SD is calculate for a "normal" curve that isn't fully symmetrical around the mean?)

Anyway, I know you aren't purporting to "solve" the whole issue in this post. But it seems to me that the numbers don't necessarily support Cremieux's claim that the statistics somehow "debunk" the whole "obesity epidemic" narrative.

Expand full comment
15 more comments...

No posts