The sampling variance of a u ratio is computed differently for independent samples (i.e., settings where the referent unrestricted standard deviation comes from an different sample than the range-restricted standard deviation) than for dependent samples (i.e., unrestricted samples from which a subset of individuals are selected to be in the incumbent sample).

The sampling variance for independent samples (the more common case) is:

$$var_{e}=\frac{u^{2}}{2}\left(\frac{1}{n_{i}-1}+\frac{1}{n_{a}-1}\right)$$

and the sampling variance for dependent samples is:

$$var_{e}=\frac{u^{2}}{2}\left(\frac{1}{n_{i}-1}-\frac{1}{n_{a}-1}\right)$$

where *u* is the u ratio, \(n_{i}\) is the incumbent sample size, and \(n_{a}\) is the applicant sample size.