Can you guess a size from height and weight?

Height and weight say a lot about a body, but not everything. Two men of 178 cm and 80 kg can have chests a few centimeters apart, and on Ruban Size's clothing templates one size spans 5 to 8 cm of chest. So the honest answer to "what size am I from my height and weight?" is a likely size, with the odds of the ones next to it.

This is how that estimate is made from published sources, what its error means, and where it stops working.

Where the equations come from

ANSUR 1988: chest, hips, inseam, foot

In 1988 the US Army measured 1,774 men and 2,208 women in detail: the ANSUR survey. Cheverud and colleagues published, in 1990, a regression for each body dimension from stature and weight, with the standard error of each estimate. These equations are public, and they are what Ruban Size uses for the chest (for women, taken at the fullest part of the bust), the underbust, the hips (around the buttocks), the inseam (floor to crotch) and the foot length.

They read simply. For a man, chest (mm) = 928.002 + 0.622 × weight (in tenths of a kilogram) minus 0.242 × stature (mm). At the same weight, a taller man has a slightly smaller chest; at the same height, a heavier one a larger chest. For a man of 178 cm and 80 kg, that gives a chest of 99.5 cm.

Bozeman et al. 2012: the waist

The waist comes from a second source: Bozeman and colleagues (2012), who fitted it on 4,641 adults of NHANES 2001 to 2002, aged 18 and up, with a body mass index (BMI) up to 40. The waist grows with BMI and with age, and the women's equation has its own age term from 35. The same man of 178 cm and 80 kg, aged 35, has a BMI of 25.25 and an estimated waist of 91.8 cm.

What the standard error means

Each estimate comes with a standard error: the typical distance between the estimate and the real measurement of people with that height and weight. About two people in three fall within one standard error; the rest are further out.

Now set that against a chart. On Ruban Size's women's tops template, one size is 5 to 7.5 cm of chest. A standard error of 3.4 cm is about half a size each way, so a woman whose estimate sits in the middle of M has a real chance of being S or L. On the men's T-shirt template, our 178 cm, 80 kg man has a chest in M (96 to 103 cm) and a waist in L (90 to 97 cm): M is the likely size, and L a real possibility.

That is why an estimate shouldn't be shown as a single size with a confident tick. It should be shown as chances.

Turning the error into chances

Ruban Size moves every estimated measurement together, from 2.5 standard errors below the estimate to 2.5 above in quarter steps, weights each point by the normal curve, and runs Find my size at each point. The share of points landing on each size is its chance: "Chance each size fits best: M 55% · S 24% · L 19%". Sizes with less than one chance in ten aren't listed.

When the second size has a chance of one in four or more, the answer says so: "Between S and M. Most likely M." Otherwise it is "Estimated size: M". Either way the confidence is never "high" for an estimate. On our benchmark, sizes given an 80 to 100 percent chance were right 85 times in 100. Take that figure with care: the simulated shoppers were built from the same equations, which flatters the estimate.

The limits

An estimate is only as wide as the people it was fitted on. Ruban Size refuses to guess outside them.

• Adults only. No estimate under 18. Children's charts are read by height and age, and Find my size asks for those directly.
• Height within the survey. Men 149.7 to 204.2 cm, women 142.8 to 187.0 cm.
• BMI from 16 to 40. Bozeman's sample stopped at 40.
• Weight outside the survey, BMI still in range: the estimate is given, flagged as less sure, because a straight line is being stretched past its data. Excluding these shoppers would have shut out many plus-size shoppers.
• Not on charts that have a height row. Those already ask for height, so an estimate would add nothing.

Out of range, the shopper reads "We can't estimate reliably from these values. Please check them, or measure yourself." Less sure, they read "Few people of your build were in the surveys, so this estimate is less sure."

Some limits can't be fixed with a check. ANSUR measured young, fit soldiers in 1988; older or heavier bodies carry weight differently. NHANES measures the waist at the top of the hip bone, while a dress chart often uses the natural waist, which is narrower, so the estimate can lean to the size above for women. The neck isn't estimated at all, because the men's neck equation is unreadable in the only copy of the report we could read. And the waist equation's ethnic terms aren't asked, so the reference group is used.

How Ruban Size shows it

On the Fit plan, Find my size adds a link: "Don't know your measurements? Use your height and weight". It replaces the measurement fields with sex (Female, Male, or Prefer not to say), height in cm or feet and inches, weight in kg or pounds, age, and an optional body shape: Like most people, Fuller chest, Fuller hips, Rounder middle.

"Prefer not to say" blends the women's and men's equations and is always marked less sure. A body shape answer moves that one measurement up within its known error, without inventing a new coefficient.

The answer shows the estimated size, the chances, the estimated measurements with their likely range, and one sentence: "This is an estimate, not a measurement. For the best fit, measure yourself and enter your measurements." Height, weight and age stay on the shopper's device. The estimate is on by default on Fit, at $9.99 a month or $99 a year; see the pricing page.

Measurements remain the better answer when a shopper has a tape, and Find my size offers them first, on every plan. The comparison of the two approaches is in size quiz or measurements, and the three kinds of recommender in what a size recommender is. For charts sized by height, like kids' clothes, start from the kids' size chart template.

Questions

How accurate is a clothing size from height and weight?

Expect the right size or the one next to it. The published equations have a typical error of about 3 cm on the chest and 5 cm on the waist, close to one clothing size, so a good tool shows the chances of each size.

Why does a size calculator ask for my age?

Age changes the waist at the same height and weight. The waist equation from Bozeman et al. 2012 includes it, and the estimate is only offered from 18.

Can kids' sizes be estimated from height and weight?

Not with these equations, which were fitted on adults. Kids' charts in Ruban Size are read by the child's height and age directly.

Does Ruban Size store my height and weight?

No. They are kept on your device, in your browser, and never sent to Ruban Size.