Using the discontinuation rule to reduce the effect of random guessing on parameter estimation in the item response theory

Abstract

The discontinuation rule is often used to reduce the effect of random guessing in psychological tests. It may also play the similar role in the item response theory. In this article, a Monte Carlo study was implemented to explore the feasibility of the four- and six-consecutive-zero discontinuation rules to reduce the effect of random guessing on parameter estimation in the Rasch model. The results showed that random guessing inflated estimation errors, and these discontinuation rules can reduce this effect on item-parameter estimation under the joint and marginal maximum likelihood, but can do so only for person-parameter estimation under the marginal maximum likelihood and expected a posteriori methods.