Conditional standard errors of measurement
It has long been acknowledged that the standard error of measurement “varies across the range of
examinee ability” (Qualls-Payne, 1992). The estimate of the standard error of measurement traditionally
used in classical test theory is an average figure (see Feldt 1984 for a particularly cogent demonstration).
When we have on hand a testing procedure with high-stakes outcomes, such as that encountered in the
application of mastery and certification tests, it is important to recognize that the standard error of
measurement in the vicinity of the cut score is likely to differ from the average. This concern is what is
reflected in Standard 2.14.