Examining Proficiency Classification in English Language Assessment Using Four Domain Score Estimation Methods

• In Event: Poster Session 6
  In Poster Session: Bilingualism/English as a Second Language

Abstract

The demand for using English language assessments for diagnostic purposes requires domain scores reporting. The relationship between domain scores is also informative for understanding English learning process. Based on the results from an empirical study, this research compared four domain-scoring methods using a simulated test. The Regression based (RB) method was found to be better in reporting domain scores of an English language test. It outperformed the number of correct method, the multidimensional item response theory (IRT) method and the unidimensional IRT method with respect to subscore recovery and classification accuracy. Since domain scores are relatively less reliable than the total test score, the meaningfulness of reporting domain scores, as well as the implications for test validity are discussed.

Authors

• Liyang Mao, Michigan State University

• Xuechun Zhou, Pearson Assessment & Information