ABSTRACT

Estimation and linking are required for any practical implementation of multistage testing (MST). Different types of MST designs are defined in Yan, Lewis, and von Davier (chapter 1, this volume). Although the details of the analysis in the present chapter depend on the specific MST design, the basic principles required are consistent for all designs. In published literature, item response theory (IRT) generally provides the basic framework for linking, scoring, and equating of an MST (Glas 1988). This chapter will follow this practice. In addition, to simplify exposition, discussion is confined to a onedimensional IRT model, and it is assumed that individuals rather than groups are to be assessed in a situation in which fairness considerations dictate that information external to the test performance cannot be employed in scoring.