Skip to content
Lord Sterling
Science

How uncertainty travels through a result

Archive 01350

Slide 1

A calculated area inherits uncertainty from its measurements

An area calculated from measured length and width is not exact. Uncertainty in both measurements affects the reported area.

More to Read

A measurement model states how the desired quantity depends on inputs. Writing that relationship first helps identify overlooked influences, such as thermal expansion when a dimension changes with temperature.

Slide 2

Sensitivity determines which input matters most

Sensitivity describes how strongly an output changes when one input changes. Equal input uncertainties need not have equal effects.

More to Read

For a rectangle, changing its length by a millimetre changes area more when the width is large. The physical relationship, rather than the number of decimal places, determines that influence.

Slide 3

Independent uncertainties can combine through squares

Independent standard uncertainties often combine through a square-root-of-squares calculation. Adding every uncertainty directly answers a different, more conservative question.

More to Read

For a sum of independent inputs with standard uncertainties of 3 and 4 units, the combined standard uncertainty is 5 units. This example assumes the sensitivity to each input is one.

Slide 4

Shared references can make errors move together

Two measurements made with the same reference may have correlated errors. Their uncertainty contributions cannot always be treated as independent.

More to Read

A shared scale error can partly cancel in a ratio, while reinforcing in a sum. Covariance expresses this connection mathematically; ignoring it can make an uncertainty estimate too large or too small.

Slide 5

The reported interval needs a stated meaning

A reported uncertainty needs a defined meaning and method. An interval without context does not reveal how it was constructed.

More to Read

Expanded uncertainty multiplies a combined standard uncertainty by a coverage factor. Interpreting that factor as a probability also requires assumptions about the distribution and available information; it is not automatic for every measurement.

Sources

Join Lord Sterling to explore your personal Discover feed and preserve what interests you.