A HYBRID Model for Test Speededness

Abstract

Assessing speededness by simple approaches such as counting the number of missing responses near the end of a test is often inadequate because many examinees switch to a guessing or random response strategy as the testing time limit approaches. Parameter estimation within item response theory (IRT) can be greatly impacted by speededness; thus, it is crucial to assess how much speededness a test may possess (Oshima, 1994). It is also critical to correct the item-parameter estimates that may have been affected by this end-of-test speededness. Examinees who switch to random responses at the end of the test, in terms of the underlying response processes, expose a very different behavior when responding to an item when compared to examinees who try to solve the item using their cognitive skill set. In order to account for these different types of response behaviors, a HYBRID model was proposed by Yamamoto (1989) and later extended to assess test speededness more directly (Yamamoto, 1990, 1995).