ABSTRACT
In chapter 7, we introduced a novel IRT-based score combination method to produce scale scores for tests that combine multiple-choice and con structed-response items; to use unidimensional IRT, we assumed that the tests, although they mix item types, are essentially unidimensional with re spect to the variation in proficiency that they measure. The only obvious disadvantage of the scoring system introduced in chapter 7 is the poten tially unwieldy size of the scoring tables. In this chapter, we develop ap proximate methods that replace the scoring tables with weighted linear combinations of the component scores.
