Principles of Multidimensional Adaptive Testing

Abstract

Tests used to measure individual differences are often designed to provide comprehensive information along several dimensions of knowledge, skill, or ability. For example, college entrance exams routinely provide separate scores on math and verbal dimensions. Some colleges may elect to base qualification on a compensatory model, where an applicant’s total score (math plus verbal) must exceed some specified cutoff. In this instance, the individual math and verbal scores may provide useful feedback to students and schools about strengths and weaknesses in aptitudes and curriculum. In other instances, colleges may elect to base qualification on a multiple-hurdlemodel, where the applicant’s scores on selected components must exceed separate cutoffs defined along each dimension. For example, a college may elect to have one qualification standard for math knowledge and another standard for verbal proficiency. Applicants may be required to meet one or the other, or both standards, to qualify for entrance. In all these instances, it is useful and important for the individual component scores to possess adequate psychometric properties, including sufficient precision and validity.