ABSTRACT
Missing responses occur in item response theory (IRT) applications in ways that may or may not have been intended by the examiner and may or may not be related to people’s ability θ. The mechanisms that produce missingness must be taken into account to draw172correct inferences. This chapter addresses missingness in IRT through the lens ofRubin’s (1976,1977) missing-data framework. Implications for Bayesian and likelihood estimation are discussed for (a) random assignment of items, including alternate test forms and matrix sampling, (b) targeted testing, including vertical linking, (c) adaptive testing, (d) not-reached items, (e) intentional omits, and (f) examinee choice of items.
