Making Better Tests with the Rasch Measurement Model

Abstract

This study had two aims. The first was to explain the process of using the Rasch measurement model to validate tests in an easy-to-understand way for those unfamiliar with the Rasch measurement model. The second was to validate two final exams with several shared items. The exams were given to two groups of students with slightly differing English listening proficiency. The two exams, a low-advanced and a high-advanced exam, were given to 76 and 45 Japanese university students, respectively. Each exam had 56 questions with 26 shared questions linking the two exams. After conducting a simple Rasch analysis, it was determined that up to 33 questions needed to be modified or deleted from subsequent versions of the exam. The unexpected number of recommended modifications and deletions suggests that, even for experienced teachers, the Rasch measurement model can be of tremendous value by offering greater precision in the assessment of students, as well as greater assistance in the validation of tests.