Tailored-Length Mastery Tests

Year: 1986

Author: Macinnis, Peter

Type of paper: Abstract refereed

The longer a mastery test is, the more reliable it becomes. But if mastery is a continuum, there is no risk of false conclusions from true masters or non-masters: the answer is to tailor the length of the test to the individual, using Bayesian assumptions. This leads to curious and novel effects, such as the desirability of items with a high negative discrimination.