Using a Universal Shadow-Test Assembler with Multistage Testing
Wim J. van der Linden, Qi Diao · 2016
A key feature of multistage testing (MST), as well as any other adaptive testing format, is sequential selection of the items to optimally adapt them to updates of the test taker’s ability estimate. At the same time, the selection has to meet all content, psychometric, and practical specifications for the complete test. This combination of a statistical objective with a set of constraints required to realize these specifications reveals that such formats belong to a class of problems more generally known as constrained combinatorial optimization problems. Other instances of this class are found widely throughout business, trade, and industry and include such problems as machine scheduling in manufacturing, vehicle routing in transportation, portfolio assembly in finance, and crew assignment in the airline industry. Each of these problems shares the common feature of the selection of an optimal combination of “objects” (machines, routes, test items, etc.) from a finite pool subject to a set of constraints.