From a statistical point of view, adaptive testing belongs to the class of problems of selecting a best subset of random variables. As in any other instance of this problem, adaptive testing is likely to capitalize on estimation error. An empirical example of the phenomenon is discussed, and a strategy for cross-validating results from adaptive testing is proposed. Results from a simulation study on the strategy are presented.
|Title of host publication||Essays on item response theory|
|Editors||Anne Boomsma, Marijtje A.J. van Duijn, Tom A.B. Snijders|
|Publication status||Published - 2001|
|Name||Lecture notes in statistics|