Fitting a Mixture Rasch Model to English as a Foreign Language Listening Tests: The Role of Cognitive and Background Variables in Explaining Latent Differential Item Functioning

Abstract

The present study uses a mixture Rasch model to examine latent differential item functioning in English as a foreign language listening tests. Participants (n = 250) took a listening and lexico-grammatical test and completed the metacognitive awareness listening questionnaire comprising problem solving (PS), planning and evaluation (PE), mental translation (MT), person knowledge (PK), and directed attention (DA). The listening test was subjected to MRM analysis where a two-latent class model had a sufficient fit. Next, an artificial neural network and a chi-square test were used to examine the nature of the latent classes. Class 1 comprised high-ability listeners capable of multitasking and obtained high PS, PE, and lexico-grammatical test scores but low DA, PK, and MT scores. Class 2 comprised low-ability listeners with limited multitasking skills who obtained high DA, PK, and MT scores but low scores on PS, PE, and the lexico-grammatical test. Finally, a model of listening comprehension is postulated and discussed.

Keywords:

• artificial neural network
• gender
• item response theory
• lexico-grammatical knowledge
• listening comprehension
• metacognitive strategy awareness
• mixture Rasch measurement