The specification of attribute structures and its effects on classification accuracy in diagnostic test design

Abstract

Diagnostic test has gained attention for its potentiality to produce fine-grained information about examinees. The dependency among attributes (i.e. attribute structure) is one of the most important factors affecting diagnostic test design. This article introduces four types of attribute structures and examines the effects of the attribute number, structure and level on classification accuracy and reliability. Results from the study help researchers and practitioners understand factors that affect classification when specifying attributes, and design diagnostic tests that provide accurate information about examinees.