Random-item Rasch models and explanatory extensions: A worked example using L2 vocabulary test item responses

Abstract

This paper describes the application and relevance of random-item Rasch models in second language (L2) vocabulary research and testing scenarios, aiming to increase understanding of this statistical method amongst researchers and academics working in the L2 assessment field and more broadly in applied linguistics. A step-by-step description of the links between Generalized Linear Mixed Models (GLMMs) and Rasch models is given. It is then demonstrated how random-item-random-person (RPRI) Rasch models (De Boeck, 2008) can be built within a GLMM framework, and the modelling of an explanatory extension is presented in which the role of word and item characteristics are tested as fixed effect covariates in explaining item difficulty in an L2 vocabulary test completed by Hungarian school-age learners of English.