Hacker News Reader

Comment by ianburrell | Hacker News Reader

Another example of confusing mean and median. The CDC is reporting the mean waist size. But the article talks about the “average man” and “most men” wouldn’t fit. That is the median.

Waist may be normal distribution like height which means they are close. But I think the distribution may be unbalanced with limit on low weight but none on high.

View on HN

Why speculate? _Anthropometric Reference Data for Children and Adults_, Table 20 "Waist circumference in centimeters for adult males aged 20 and over and number of examined persons, mean, standard error of the mean, and selected percentiles, by race and Hispanic origin and age: United States, 2015–2018" is right at https://www.cdc.gov/nchs/data/series/sr_03/sr03-046-508.pdf

The mean in the very first row is 102.9 cm (the 40.5 inches in the OP), while the 50th percentile is actually lower at 101.8 (39.7 inches).

Their point was that mean might be higher than median. That said, from your data is isn’t much higher, so they’re broadly wrong. Headline fairly close to accurate for median too

I don't see an issue. The headline says average and that normally means mean, not median. It also doesn't say that half of people are excluded from it.

People reading the article are used to hearing about the average household credit card debt in the USA, and that's also the mean.

Edit: I missed the most men part. Other than that it's fine. That would be like if those credit card debt stories said that "most households have more than $7000 in credit card debt" if the average household credit card debt was found to be $7,486 like it was earlier this year: https://thehill.com/finance/3821799-a-growing-number-of-amer.... That would sound wrong to me, and it is.

That’s exactly it. The average gets pulled out of proportion by outliers, so there’s a good chance that far more than 50% of US American man can ride it.

It isn't getting skewed, it is simply being different than the median. Very common in statistics.

What?

Expecting a gaussian distribution everywhere. For instance in dead pixels in monitors. Say that most have no dead pixels but there are 1000 with one dead pixel. Should there be 1000 with an extra pixel?

I don’t understand what you are saying. The article says 50% of men. I said with the average, there is no need for it to be 50%, with the median there would be. I have no idea how what you say is related to that.

> The average gets pulled out of proportion by outliers

This places a value judgment on the mean not being close to the median

No, it simply says there’s a chance. That is what this whole subthread was about, and you argued against it for some inexplicable reason. But now I see you even edited your original comment to say you misunderstood, so I’m thinking you are just arguing for the sake of it.

The article says:

“According to the CDC, the average American man's waist is 40.5 inches, which means most American men would be excluded from the ride.”

If “average” means “mean”, then clearly this reasoning is broken.

Ah, they expanded poorly on it. They also say that again a paragraph above that. Chalk one up for https://norvig.com/reporters-and-parrots.html Maybe ChatGPT is already capable of replacing human parroting...

No, whether average refers to mean or median depends on the context.

Distributions are available here:

https://www.cdc.gov/nchs/data/series/sr_03/sr03-046-508.pdf

Median male adult weight: 192.6 pounds

Yes, this keeps coming up on HN. It's a bimodal distribution, so if 85% of the population can ride on your ride, it's still a solid business proposition.

Reply on news.ycombinator.com