Comparing Surface and Underlying Local Dependence Levels via Polychoric Correlations.

Item response theory (IRT) is a set of psychometric models used in the social and behavioral sciences. As part of applying these models in practice, a number of assumptions are made. A large literature exists assessing the extent to which these assumptions are satisfied in a given data set. One of these assumptions, local independence, is the focus of the research described here. When the local independence assumption is violated, there is said to be local dependence (LD). Several different models of LD have been proposed, and a number of studies have been conducted examining the performance of different methods at detecting LD. Underlying LD (ULD) and surface LD (SLD) were proposed as two possible mechanisms underlying observed LD in an early exploration of detection procedures. In a number of previous studies, it appears as though ULD is more difficult to detect than SLD. In this article, the authors demonstrate a procedure to examine comparability of induced LD and present results, which suggest a re-interpretation of existing studies on LD detection.