| Abstract: | In an article in the Winter 2011 issue of the Journal of Educational Measurement, van der Linden, Jeon, and Ferrara suggested that “test takers should trust their initial instincts and retain their initial responses when they have the opportunity to review test items.” They presented a complex IRT model that appeared to show that students would be worse off by changing answers. As noted in a subsequent erratum, this conclusion was based on flawed data, and that the correct data could not be analyzed by their method because the model failed to converge. This left their basic question on the value of answer changing unanswered. A much more direct approach is to simply count the number of examinees whose scores after an opportunity to change answers are higher, lower, or the same as their initial scores. Using the same data set as the original article, an overwhelming majority of the students received higher scores after the opportunity to change answers. |
