Equating and Linking of Performance Assessments

Abstract

Performance assessments (PA) are used in various contexts of large-scale educational assessment. It is often desirable to compare examinee performance on different forms of an assessment or on the same forms administered at different times. An overview of linking methods applied to PA is presented: major issues and recent developments in linking PAs are discussed, three common linking designs (single group, randomly equivalent groups, and nonequivalent groups with anchor items) are compared, and two major linking methodologies [classical and item response theory (IRT)] are evaluated from the PA perspective. Also described are two classical equating methods (linear and equipercentile) and several IRT equating methods (item response function, vertical, common population, and multiple-group). Areas for future research are identified.