ABSTRACT

Assigning examinees to proficiency categories is a practice with a long history in testing. By definition, such assignments are classifications. Recent psychometric advances have yielded models that use classification approaches to scaling and assessing examinee ability. Models such as mixture Rasch models (i.e., Rost, 1990), diagnostic classification models (i.e., Rupp & Templin, 2008), and more general latent class models (i.e., Lazarsfeld & Henry, 1968) all are well suited for use in standard setting and proficiency assignment. Their use provides greater transparency in delineating the statistical criteria for attaining subject matter proficiency and, perhaps more importantly, allows for a reduced level of classification error when compared with more traditional approaches, such as the use of cut scores or thresholds.