Methods and sources

How these positions are worked out

A result has to hold up for two readers: a parent deciding whether to believe it, and a professional checking it. What follows is written to the second one’s standard, in words the first can read.

One point, two points and a line are not the same claim

  • One measurement

    Needs
    Age, sex, one measurement
    Can say
    Where a child sits among peers today — which percentile, and what the median is.
    Cannot say
    How fast they are growing. The same 50th percentile may be one a child has sat at all along, or one they have just dropped to from the 90th.

  • The change between two

    Needs
    Two dated measurements taken the same way
    Can say
    How far apart, how much was gained, how far the position moved.
    Cannot say
    A direction. Two points make a line with one direction, and that direction may be nothing but one measurement’s error.

  • A trend

    Needs
    Three or more, across enough time
    Can say
    Whether the position is moving steadily.
    Cannot say
    Anything here. It needs past records kept, and nothing here is kept.

  • A projected range

    Needs
    Assumptions stated in the open
    Can say
    Roughly which range, while those assumptions hold. The adult height range is the only place this happens.
    Cannot say
    A single figure. Giving one would claim more than the method can support.

Which table each figure came from

Every standard on file records who published it, in what year, from whose children, and whether it measures how healthy children should grow or how one cohort actually grew — a difference that changes how the same percentile reads. Every figure on a result page links back to the table it came from.

Open the library →

How the numbers are produced

  • LMS parameters come straight from each standard’s published tables, with no refitting. Between published rows, L, M and S are each interpolated, and the page marks the result as derived.
  • Standards published as cut-offs — China’s national standard among them — give only the percentile lines they printed, never extrapolated past them.
  • Where a region’s table does not reach an age, the lookup follows that region’s own routing to another standard, and the page names whose table it landed on rather than passing it off as local.
  • The second birthday is where the method changes: recumbent length before, standing height after. Two measurements straddling it show a drop of about 0.67 cm that is the ruler changing, not the child shrinking.
  • Preterm infants are looked up by corrected age, with the correction ending on the RCPCH convention — and the page says that is a convention, not the standard’s own rule.

When no result is given

  • The age falls outside what that standard published — said plainly, never extrapolated.
  • The indicator stops meaning anything at that age: weight-for-age past ten cannot tell tall from heavy, so the page points to BMI instead.
  • Two measurements less than 90 days apart — most of that difference is measurement error, and annualising it produces a confident-looking figure made of noise.
  • The two were not measured the same way, or there is no way to tell (no date of birth).
  • Vision, hearing, teeth and blood counts have no checkable child reference tables here, so nothing at all is said about them.