3 ms·

The absence of evidence is not the evidence of absence. Especially when the absence of evidence is simply a report that is not looking for what you're looking f

The absence of evidence is not the evidence of absence. Especially when the absence of evidence is simply a report that is not looking for what you're looking for. Either way, BMI clearly has some use. Or do you think this chart is completely random and devoid of correlation? http://upload.wikimedia.org/wikipedia/commons/b/b7/Correlation_between_BMI_and_Percent_Body_Fat_for_Men_in_NCHS%27_NHANES_1994_Data.PNG http://upload.wikimedia.org/wikipedia/commons/b/b7/Correlati... Do you think the ordering of countries on the BBC link is purely random and has nothing to do with how many fat people are in each country?

dredmorbius 14y ago
That chart is largely irrelevant to the discussion in several ways. Your question is specious. It shows population correlations. In a general population, where a range of bodytypes may be assumed, there will be a correlation between BMI and body fat percentage. There's relatively little argument over that, though as I mentioned above, the actual utility of BMI for epidemiological / population studies is being question. - The question isn't "is there a correlation?", but "is the correlation between BMI and body fat stronger or weaker than for other estimators of body fat?". For which you'd want to look at the r^2, RSS (residual sum of squares), and perhaps plotted residuals to determine the appropriateness of the model, in comparison with other estimators: waist circumference, height/waist ratio (most people also have a pretty good idea of their waist size), circumference and caliper estimators, etc. - To paraphrase Donald Rumsfeld, in epidemiological population studies, you conduct research with the data you have, not with the data you want. Height and weight are widely collected and reported, even from sources such as DMV files (though they may also be wildly inaccurate). For large-scale or national studies, this makes for large datasets. The relevance and usefulness of the data may not be as high as, say, DEXA analysis, or even waist to height ratio (http://www.webmd.com/heart-disease/news/20050606/waist-height-ratio-may-show-heart-disease-risk http://www.webmd.com/heart-disease/news/20050606/waist-heigh...), but it's there. And for at least some purposes, useful. Extending the use to individuals, where in most settings (home, gym, doctor's office, clinic, health fair), availability of necessary measurement tools (tape measure, calipers, even the wildly inaccurate handheld electronic bioimpedance monitors) is readily facilitated, there's really no call to rely on BMI. - The problem with BMI is that it's essentially saying "if you fit our model of body fat relative to height and weight, then here is where you fall on our model of obesity". Again, for a general population this relationship holds reasonably well, but for an individual, where the question specifically is "what is my body fat percentage", BMI is the wrong tool for the job. It fails notably for specific populations of interest -- it will consistently underreport obesity in less-healthy populations, and overreports obesity in more-healthy populations. It can also result in nonsensical intervention recommendations: most unfit people will benefit by efforts to increase lean muscle mass, despite the fact that for those with lower BMIs this will decrease their putative health score by increasing overall body mass. I suspect this is among the reasons strength training is under-recommended for increasing fitness (most literature addresses "diet" and "exercise" generally, but makes little or no distinction between cardio and strength training, or focuses virtually exclusively on cardio).
- streptomycin 14y ago
  > The question isn't "is there a correlation?", but "is the correlation between BMI and body fat stronger or weaker than for other estimators of body fat?". As I've been repeatedly saying, the other estimators of body fat are much more difficult to calculate. Most people already know their height and weight. Most people don't know their waist circumference and have absolutely no idea how to calculate BF%. That's why sites like this use BMI, so that their visitors can gain some information rather than no information.
  - dredmorbius 14y ago
    I'll simply note a very strong disagreement that people don't know, or can't readily obtain, their waist measurement. Particularly in a setting at which they'd be particularly interested in some measure of general fitness: doctor's office, clinic, gym, health fair, etc. Which means that your statement that other measures are harder to obtain or less accurate is simply false. Among the problems of BMI is that it is official health policy. It's the definition of obesity according to the US government. It's the basis of insurance coverage / premium assessments and other abuses. Changing policy would mean that doctors' offices would measure your height and waist, to give a far more accurate fitness assessment. Over time, epidemiologists would have this dataset generally available. Really. It's long past time to retire BMI. Why are you so wedded to it?

dredmorbius 14y ago

That chart is largely irrelevant to the discussion in several ways. Your question is specious. It shows population correlations. In a general population, where a range of bodytypes may be assumed, there will be a correlation between BMI and body fat percentage. There's relatively little argument over that, though as I mentioned above, the actual utility of BMI for epidemiological / population studies is being question. - The question isn't "is there a correlation?", but "is the correlation between BMI and body fat stronger or weaker than for other estimators of body fat?". For which you'd want to look at the r^2, RSS (residual sum of squares), and perhaps plotted residuals to determine the appropriateness of the model, in comparison with other estimators: waist circumference, height/waist ratio (most people also have a pretty good idea of their waist size), circumference and caliper estimators, etc. - To paraphrase Donald Rumsfeld, in epidemiological population studies, you conduct research with the data you have, not with the data you want. Height and weight are widely collected and reported, even from sources such as DMV files (though they may also be wildly inaccurate). For large-scale or national studies, this makes for large datasets. The relevance and usefulness of the data may not be as high as, say, DEXA analysis, or even waist to height ratio (http://www.webmd.com/heart-disease/news/20050606/waist-height-ratio-may-show-heart-disease-risk http://www.webmd.com/heart-disease/news/20050606/waist-heigh...), but it's there. And for at least some purposes, useful. Extending the use to individuals, where in most settings (home, gym, doctor's office, clinic, health fair), availability of necessary measurement tools (tape measure, calipers, even the wildly inaccurate handheld electronic bioimpedance monitors) is readily facilitated, there's really no call to rely on BMI. - The problem with BMI is that it's essentially saying "if you fit our model of body fat relative to height and weight, then here is where you fall on our model of obesity". Again, for a general population this relationship holds reasonably well, but for an individual, where the question specifically is "what is my body fat percentage", BMI is the wrong tool for the job. It fails notably for specific populations of interest -- it will consistently underreport obesity in less-healthy populations, and overreports obesity in more-healthy populations. It can also result in nonsensical intervention recommendations: most unfit people will benefit by efforts to increase lean muscle mass, despite the fact that for those with lower BMIs this will decrease their putative health score by increasing overall body mass. I suspect this is among the reasons strength training is under-recommended for increasing fitness (most literature addresses "diet" and "exercise" generally, but makes little or no distinction between cardio and strength training, or focuses virtually exclusively on cardio).

streptomycin 14y ago
> The question isn't "is there a correlation?", but "is the correlation between BMI and body fat stronger or weaker than for other estimators of body fat?". As I've been repeatedly saying, the other estimators of body fat are much more difficult to calculate. Most people already know their height and weight. Most people don't know their waist circumference and have absolutely no idea how to calculate BF%. That's why sites like this use BMI, so that their visitors can gain some information rather than no information.
- dredmorbius 14y ago
  I'll simply note a very strong disagreement that people don't know, or can't readily obtain, their waist measurement. Particularly in a setting at which they'd be particularly interested in some measure of general fitness: doctor's office, clinic, gym, health fair, etc. Which means that your statement that other measures are harder to obtain or less accurate is simply false. Among the problems of BMI is that it is official health policy. It's the definition of obesity according to the US government. It's the basis of insurance coverage / premium assessments and other abuses. Changing policy would mean that doctors' offices would measure your height and waist, to give a far more accurate fitness assessment. Over time, epidemiologists would have this dataset generally available. Really. It's long past time to retire BMI. Why are you so wedded to it?