 Coverage factor

A coverage factor $k$ multiplies combined standard uncertainty to give expanded uncertainty, $U=k{u}_c$. It is dimensionless. Choosing it requires a target coverage and assumptions about the result's distribution and the reliability of the estimated standard deviation.

For a normal distribution with known standard deviation, $k=1.96$ gives a central interval of about 95 percent probability. With a small sample and an estimated variance, an appropriate Student's t-distribution quantile can be larger. A stated factor therefore communicates a method, not merely an error-bar width.^[ref]