Testing the Measurement Invariance of Large-Scale Assessment of Foreign Language Proficiency Using Multiple Group Covariance Structure Analyses and Item Response Modeling

138 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2008.This study has addressed whether the Test of English as a Foreign Language (TOEFL), which is argued to evaluate the potential success of non-native speakers of English to understand and use standard American English at English speaking higher education institutions, measures the same abilities or attributes to the same degree across the different first language (L1) groups in a latent variable model. This dissertation has the following characteristics: Measurement invariance has been addressed through the methodological sophistication of both simultaneous multiple group covariance structure analyses and item response theory (IRT) modeling as research paradigms, ultimately contributing to the validation of a standardized large-scale foreign language proficiency test for non-native speakers of English; Dimensionality of large-scale English proficiency test for non-native applicants has been investigated to identify the internal latent structure of the test by using the confirmatory factor analysis (CFA) model-testing procedures and IRT model-fitting approaches while testing both full and partial measurement invariance models, addressing construct validity of the TOEFL for qualitatively different groups of examinees. Further, this study has explored the relation between CFA and IRT models and their differential utility for representing data and testing theoretical hypotheses of score comparability, which ensures that the meaning and interpretation of the test score are comparable on the same measurement scale for different ethnic groups of examinees of the test. The present study has focused on factor structure through CFA and invariance characteristics of item parameters through IRT analyses to test measurement invariance. An interpretation of the findings and practical implications for the future use of this particular test instrument across different L1 backgrounds are discussed.U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD