A Development of The 3PL MMM-IRT Model for Identifying Latent Class

Abstract

While multilevel mixed-measurement IRT is crucial for identifying latent class, the two-parameter IRT Model used in previous studies is inadequate. The three-parameter is therefore suggested to improve such identification. How can a three parameters multilevel mixed-measurement IRT model be applied to improve the identification, classification, and measurement of school latent class? This research aims to improve the effectiveness of the model for identifying latent class by proposing the three-parameter multilevel mixed-measurement IRT Model (3PL MMM-IRT Model). This study argues that the developed model does not only enhance efficiency of parameter estimation, but is also suitable for using in school context. The proposed model was developed by using Program R. Students’ scores from Thailand's annual Ordinary National Educational Test (O-NET) were used. Four hundred and sixteen schools were randomly selected into this study. Students’ abilities were first estimated by 3pl-IRT model. Later, abilities of schools were calculated from their students’ abilities. Finally, school latent classes were reached by using latent class model. The results indicated that the developed model improves the effectiveness of the classification and measurement of school latent class. The findings also enable schools and teachers to identify not only varied abilities but also strengths and weaknesses of their students, which are crucial for improving their teaching procedure and class management. This study reveals that the inclusion of guessing parameters into the Multilevel Mixture IRT model helps to improve the effectiveness of such model. Furthermore, the developed model can be applied to classify school latent class based on students’ abilities, which contributes to increased teaching performance and effectiveness.